OSDN Git Service

b69a733b5b42d96443b36ca508d868b17c126c76
[uclinux-h8/linux.git] / fs / ext4 / ext4.h
1 /*
2  *  ext4.h
3  *
4  * Copyright (C) 1992, 1993, 1994, 1995
5  * Remy Card (card@masi.ibp.fr)
6  * Laboratoire MASI - Institut Blaise Pascal
7  * Universite Pierre et Marie Curie (Paris VI)
8  *
9  *  from
10  *
11  *  linux/include/linux/minix_fs.h
12  *
13  *  Copyright (C) 1991, 1992  Linus Torvalds
14  */
15
16 #ifndef _EXT4_H
17 #define _EXT4_H
18
19 #include <linux/types.h>
20 #include <linux/blkdev.h>
21 #include <linux/magic.h>
22 #include <linux/jbd2.h>
23 #include <linux/quota.h>
24 #include <linux/rwsem.h>
25 #include <linux/rbtree.h>
26 #include <linux/seqlock.h>
27 #include <linux/mutex.h>
28 #include <linux/timer.h>
29 #include <linux/wait.h>
30 #include <linux/blockgroup_lock.h>
31 #include <linux/percpu_counter.h>
32 #include <crypto/hash.h>
33 #ifdef __KERNEL__
34 #include <linux/compat.h>
35 #endif
36
37 /*
38  * The fourth extended filesystem constants/structures
39  */
40
41 /*
42  * Define EXT4FS_DEBUG to produce debug messages
43  */
44 #undef EXT4FS_DEBUG
45
46 /*
47  * Debug code
48  */
49 #ifdef EXT4FS_DEBUG
50 #define ext4_debug(f, a...)                                             \
51         do {                                                            \
52                 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:",        \
53                         __FILE__, __LINE__, __func__);                  \
54                 printk(KERN_DEBUG f, ## a);                             \
55         } while (0)
56 #else
57 #define ext4_debug(fmt, ...)    no_printk(fmt, ##__VA_ARGS__)
58 #endif
59
60 /*
61  * Turn on EXT_DEBUG to get lots of info about extents operations.
62  */
63 #define EXT_DEBUG__
64 #ifdef EXT_DEBUG
65 #define ext_debug(fmt, ...)     printk(fmt, ##__VA_ARGS__)
66 #else
67 #define ext_debug(fmt, ...)     no_printk(fmt, ##__VA_ARGS__)
68 #endif
69
70 #define EXT4_ERROR_INODE(inode, fmt, a...) \
71         ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a)
72
73 #define EXT4_ERROR_INODE_BLOCK(inode, block, fmt, a...)                 \
74         ext4_error_inode((inode), __func__, __LINE__, (block), (fmt), ## a)
75
76 #define EXT4_ERROR_FILE(file, block, fmt, a...)                         \
77         ext4_error_file((file), __func__, __LINE__, (block), (fmt), ## a)
78
79 /* data type for block offset of block group */
80 typedef int ext4_grpblk_t;
81
82 /* data type for filesystem-wide blocks number */
83 typedef unsigned long long ext4_fsblk_t;
84
85 /* data type for file logical block number */
86 typedef __u32 ext4_lblk_t;
87
88 /* data type for block group number */
89 typedef unsigned int ext4_group_t;
90
91 /*
92  * Flags used in mballoc's allocation_context flags field.
93  *
94  * Also used to show what's going on for debugging purposes when the
95  * flag field is exported via the traceport interface
96  */
97
98 /* prefer goal again. length */
99 #define EXT4_MB_HINT_MERGE              0x0001
100 /* blocks already reserved */
101 #define EXT4_MB_HINT_RESERVED           0x0002
102 /* metadata is being allocated */
103 #define EXT4_MB_HINT_METADATA           0x0004
104 /* first blocks in the file */
105 #define EXT4_MB_HINT_FIRST              0x0008
106 /* search for the best chunk */
107 #define EXT4_MB_HINT_BEST               0x0010
108 /* data is being allocated */
109 #define EXT4_MB_HINT_DATA               0x0020
110 /* don't preallocate (for tails) */
111 #define EXT4_MB_HINT_NOPREALLOC         0x0040
112 /* allocate for locality group */
113 #define EXT4_MB_HINT_GROUP_ALLOC        0x0080
114 /* allocate goal blocks or none */
115 #define EXT4_MB_HINT_GOAL_ONLY          0x0100
116 /* goal is meaningful */
117 #define EXT4_MB_HINT_TRY_GOAL           0x0200
118 /* blocks already pre-reserved by delayed allocation */
119 #define EXT4_MB_DELALLOC_RESERVED       0x0400
120 /* We are doing stream allocation */
121 #define EXT4_MB_STREAM_ALLOC            0x0800
122 /* Use reserved root blocks if needed */
123 #define EXT4_MB_USE_ROOT_BLOCKS         0x1000
124 /* Use blocks from reserved pool */
125 #define EXT4_MB_USE_RESERVED            0x2000
126
127 struct ext4_allocation_request {
128         /* target inode for block we're allocating */
129         struct inode *inode;
130         /* how many blocks we want to allocate */
131         unsigned int len;
132         /* logical block in target inode */
133         ext4_lblk_t logical;
134         /* the closest logical allocated block to the left */
135         ext4_lblk_t lleft;
136         /* the closest logical allocated block to the right */
137         ext4_lblk_t lright;
138         /* phys. target (a hint) */
139         ext4_fsblk_t goal;
140         /* phys. block for the closest logical allocated block to the left */
141         ext4_fsblk_t pleft;
142         /* phys. block for the closest logical allocated block to the right */
143         ext4_fsblk_t pright;
144         /* flags. see above EXT4_MB_HINT_* */
145         unsigned int flags;
146 };
147
148 /*
149  * Logical to physical block mapping, used by ext4_map_blocks()
150  *
151  * This structure is used to pass requests into ext4_map_blocks() as
152  * well as to store the information returned by ext4_map_blocks().  It
153  * takes less room on the stack than a struct buffer_head.
154  */
155 #define EXT4_MAP_NEW            (1 << BH_New)
156 #define EXT4_MAP_MAPPED         (1 << BH_Mapped)
157 #define EXT4_MAP_UNWRITTEN      (1 << BH_Unwritten)
158 #define EXT4_MAP_BOUNDARY       (1 << BH_Boundary)
159 #define EXT4_MAP_UNINIT         (1 << BH_Uninit)
160 /* Sometimes (in the bigalloc case, from ext4_da_get_block_prep) the caller of
161  * ext4_map_blocks wants to know whether or not the underlying cluster has
162  * already been accounted for. EXT4_MAP_FROM_CLUSTER conveys to the caller that
163  * the requested mapping was from previously mapped (or delayed allocated)
164  * cluster. We use BH_AllocFromCluster only for this flag. BH_AllocFromCluster
165  * should never appear on buffer_head's state flags.
166  */
167 #define EXT4_MAP_FROM_CLUSTER   (1 << BH_AllocFromCluster)
168 #define EXT4_MAP_FLAGS          (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
169                                  EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY |\
170                                  EXT4_MAP_UNINIT | EXT4_MAP_FROM_CLUSTER)
171
172 struct ext4_map_blocks {
173         ext4_fsblk_t m_pblk;
174         ext4_lblk_t m_lblk;
175         unsigned int m_len;
176         unsigned int m_flags;
177 };
178
179 /*
180  * Flags for ext4_io_end->flags
181  */
182 #define EXT4_IO_END_UNWRITTEN   0x0001
183 #define EXT4_IO_END_ERROR       0x0002
184 #define EXT4_IO_END_DIRECT      0x0004
185
186 /*
187  * For converting uninitialized extents on a work queue. 'handle' is used for
188  * buffered writeback.
189  */
190 typedef struct ext4_io_end {
191         struct list_head        list;           /* per-file finished IO list */
192         handle_t                *handle;        /* handle reserved for extent
193                                                  * conversion */
194         struct inode            *inode;         /* file being written to */
195         unsigned int            flag;           /* unwritten or not */
196         loff_t                  offset;         /* offset in the file */
197         ssize_t                 size;           /* size of the extent */
198         struct kiocb            *iocb;          /* iocb struct for AIO */
199         int                     result;         /* error value for AIO */
200         atomic_t                count;          /* reference counter */
201 } ext4_io_end_t;
202
203 struct ext4_io_submit {
204         int                     io_op;
205         struct bio              *io_bio;
206         ext4_io_end_t           *io_end;
207         sector_t                io_next_block;
208 };
209
210 /*
211  * Special inodes numbers
212  */
213 #define EXT4_BAD_INO             1      /* Bad blocks inode */
214 #define EXT4_ROOT_INO            2      /* Root inode */
215 #define EXT4_USR_QUOTA_INO       3      /* User quota inode */
216 #define EXT4_GRP_QUOTA_INO       4      /* Group quota inode */
217 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
218 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
219 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
220 #define EXT4_JOURNAL_INO         8      /* Journal inode */
221
222 /* First non-reserved inode for old ext4 filesystems */
223 #define EXT4_GOOD_OLD_FIRST_INO 11
224
225 /*
226  * Maximal count of links to a file
227  */
228 #define EXT4_LINK_MAX           65000
229
230 /*
231  * Macro-instructions used to manage several block sizes
232  */
233 #define EXT4_MIN_BLOCK_SIZE             1024
234 #define EXT4_MAX_BLOCK_SIZE             65536
235 #define EXT4_MIN_BLOCK_LOG_SIZE         10
236 #define EXT4_MAX_BLOCK_LOG_SIZE         16
237 #ifdef __KERNEL__
238 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
239 #else
240 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
241 #endif
242 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
243 #define EXT4_CLUSTER_SIZE(s)            (EXT4_BLOCK_SIZE(s) << \
244                                          EXT4_SB(s)->s_cluster_bits)
245 #ifdef __KERNEL__
246 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
247 # define EXT4_CLUSTER_BITS(s)           (EXT4_SB(s)->s_cluster_bits)
248 #else
249 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
250 #endif
251 #ifdef __KERNEL__
252 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
253 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
254 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
255 #else
256 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
257                                  EXT4_GOOD_OLD_INODE_SIZE : \
258                                  (s)->s_inode_size)
259 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
260                                  EXT4_GOOD_OLD_FIRST_INO : \
261                                  (s)->s_first_ino)
262 #endif
263 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
264
265 /* Translate a block number to a cluster number */
266 #define EXT4_B2C(sbi, blk)      ((blk) >> (sbi)->s_cluster_bits)
267 /* Translate a cluster number to a block number */
268 #define EXT4_C2B(sbi, cluster)  ((cluster) << (sbi)->s_cluster_bits)
269 /* Translate # of blks to # of clusters */
270 #define EXT4_NUM_B2C(sbi, blks) (((blks) + (sbi)->s_cluster_ratio - 1) >> \
271                                  (sbi)->s_cluster_bits)
272
273 /*
274  * Structure of a blocks group descriptor
275  */
276 struct ext4_group_desc
277 {
278         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
279         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
280         __le32  bg_inode_table_lo;      /* Inodes table block */
281         __le16  bg_free_blocks_count_lo;/* Free blocks count */
282         __le16  bg_free_inodes_count_lo;/* Free inodes count */
283         __le16  bg_used_dirs_count_lo;  /* Directories count */
284         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
285         __le32  bg_exclude_bitmap_lo;   /* Exclude bitmap for snapshots */
286         __le16  bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bbitmap) LE */
287         __le16  bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+ibitmap) LE */
288         __le16  bg_itable_unused_lo;    /* Unused inodes count */
289         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
290         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
291         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
292         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
293         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
294         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
295         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
296         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
297         __le32  bg_exclude_bitmap_hi;   /* Exclude bitmap block MSB */
298         __le16  bg_block_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bbitmap) BE */
299         __le16  bg_inode_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+ibitmap) BE */
300         __u32   bg_reserved;
301 };
302
303 #define EXT4_BG_INODE_BITMAP_CSUM_HI_END        \
304         (offsetof(struct ext4_group_desc, bg_inode_bitmap_csum_hi) + \
305          sizeof(__le16))
306 #define EXT4_BG_BLOCK_BITMAP_CSUM_HI_END        \
307         (offsetof(struct ext4_group_desc, bg_block_bitmap_csum_hi) + \
308          sizeof(__le16))
309
310 /*
311  * Structure of a flex block group info
312  */
313
314 struct flex_groups {
315         atomic64_t      free_clusters;
316         atomic_t        free_inodes;
317         atomic_t        used_dirs;
318 };
319
320 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
321 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
322 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
323
324 /*
325  * Macro-instructions used to manage group descriptors
326  */
327 #define EXT4_MIN_DESC_SIZE              32
328 #define EXT4_MIN_DESC_SIZE_64BIT        64
329 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
330 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
331 #ifdef __KERNEL__
332 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
333 # define EXT4_CLUSTERS_PER_GROUP(s)     (EXT4_SB(s)->s_clusters_per_group)
334 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
335 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
336 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
337 #else
338 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
339 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
340 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
341 #endif
342
343 /*
344  * Constants relative to the data blocks
345  */
346 #define EXT4_NDIR_BLOCKS                12
347 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
348 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
349 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
350 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
351
352 /*
353  * Inode flags
354  */
355 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
356 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
357 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
358 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
359 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
360 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
361 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
362 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
363 /* Reserved for compression usage... */
364 #define EXT4_DIRTY_FL                   0x00000100
365 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
366 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
367 #define EXT4_ECOMPR_FL                  0x00000800 /* Compression error */
368 /* End compression flags --- maybe not all used */
369 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
370 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
371 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
372 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
373 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
374 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
375 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
376 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
377 #define EXT4_EA_INODE_FL                0x00200000 /* Inode used for large EA */
378 #define EXT4_EOFBLOCKS_FL               0x00400000 /* Blocks allocated beyond EOF */
379 #define EXT4_INLINE_DATA_FL             0x10000000 /* Inode has inline data. */
380 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
381
382 #define EXT4_FL_USER_VISIBLE            0x004BDFFF /* User visible flags */
383 #define EXT4_FL_USER_MODIFIABLE         0x004380FF /* User modifiable flags */
384
385 /* Flags that should be inherited by new inodes from their parent. */
386 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
387                            EXT4_SYNC_FL | EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
388                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
389                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL)
390
391 /* Flags that are appropriate for regular files (all but dir-specific ones). */
392 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
393
394 /* Flags that are appropriate for non-directories/regular files. */
395 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
396
397 /* Mask out flags that are inappropriate for the given type of inode. */
398 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
399 {
400         if (S_ISDIR(mode))
401                 return flags;
402         else if (S_ISREG(mode))
403                 return flags & EXT4_REG_FLMASK;
404         else
405                 return flags & EXT4_OTHER_FLMASK;
406 }
407
408 /*
409  * Inode flags used for atomic set/get
410  */
411 enum {
412         EXT4_INODE_SECRM        = 0,    /* Secure deletion */
413         EXT4_INODE_UNRM         = 1,    /* Undelete */
414         EXT4_INODE_COMPR        = 2,    /* Compress file */
415         EXT4_INODE_SYNC         = 3,    /* Synchronous updates */
416         EXT4_INODE_IMMUTABLE    = 4,    /* Immutable file */
417         EXT4_INODE_APPEND       = 5,    /* writes to file may only append */
418         EXT4_INODE_NODUMP       = 6,    /* do not dump file */
419         EXT4_INODE_NOATIME      = 7,    /* do not update atime */
420 /* Reserved for compression usage... */
421         EXT4_INODE_DIRTY        = 8,
422         EXT4_INODE_COMPRBLK     = 9,    /* One or more compressed clusters */
423         EXT4_INODE_NOCOMPR      = 10,   /* Don't compress */
424         EXT4_INODE_ECOMPR       = 11,   /* Compression error */
425 /* End compression flags --- maybe not all used */
426         EXT4_INODE_INDEX        = 12,   /* hash-indexed directory */
427         EXT4_INODE_IMAGIC       = 13,   /* AFS directory */
428         EXT4_INODE_JOURNAL_DATA = 14,   /* file data should be journaled */
429         EXT4_INODE_NOTAIL       = 15,   /* file tail should not be merged */
430         EXT4_INODE_DIRSYNC      = 16,   /* dirsync behaviour (directories only) */
431         EXT4_INODE_TOPDIR       = 17,   /* Top of directory hierarchies*/
432         EXT4_INODE_HUGE_FILE    = 18,   /* Set to each huge file */
433         EXT4_INODE_EXTENTS      = 19,   /* Inode uses extents */
434         EXT4_INODE_EA_INODE     = 21,   /* Inode used for large EA */
435         EXT4_INODE_EOFBLOCKS    = 22,   /* Blocks allocated beyond EOF */
436         EXT4_INODE_INLINE_DATA  = 28,   /* Data in inode. */
437         EXT4_INODE_RESERVED     = 31,   /* reserved for ext4 lib */
438 };
439
440 /*
441  * Since it's pretty easy to mix up bit numbers and hex values, we use a
442  * build-time check to make sure that EXT4_XXX_FL is consistent with respect to
443  * EXT4_INODE_XXX. If all is well, the macros will be dropped, so, it won't cost
444  * any extra space in the compiled kernel image, otherwise, the build will fail.
445  * It's important that these values are the same, since we are using
446  * EXT4_INODE_XXX to test for flag values, but EXT4_XXX_FL must be consistent
447  * with the values of FS_XXX_FL defined in include/linux/fs.h and the on-disk
448  * values found in ext2, ext3 and ext4 filesystems, and of course the values
449  * defined in e2fsprogs.
450  *
451  * It's not paranoia if the Murphy's Law really *is* out to get you.  :-)
452  */
453 #define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG))
454 #define CHECK_FLAG_VALUE(FLAG) BUILD_BUG_ON(!TEST_FLAG_VALUE(FLAG))
455
456 static inline void ext4_check_flag_values(void)
457 {
458         CHECK_FLAG_VALUE(SECRM);
459         CHECK_FLAG_VALUE(UNRM);
460         CHECK_FLAG_VALUE(COMPR);
461         CHECK_FLAG_VALUE(SYNC);
462         CHECK_FLAG_VALUE(IMMUTABLE);
463         CHECK_FLAG_VALUE(APPEND);
464         CHECK_FLAG_VALUE(NODUMP);
465         CHECK_FLAG_VALUE(NOATIME);
466         CHECK_FLAG_VALUE(DIRTY);
467         CHECK_FLAG_VALUE(COMPRBLK);
468         CHECK_FLAG_VALUE(NOCOMPR);
469         CHECK_FLAG_VALUE(ECOMPR);
470         CHECK_FLAG_VALUE(INDEX);
471         CHECK_FLAG_VALUE(IMAGIC);
472         CHECK_FLAG_VALUE(JOURNAL_DATA);
473         CHECK_FLAG_VALUE(NOTAIL);
474         CHECK_FLAG_VALUE(DIRSYNC);
475         CHECK_FLAG_VALUE(TOPDIR);
476         CHECK_FLAG_VALUE(HUGE_FILE);
477         CHECK_FLAG_VALUE(EXTENTS);
478         CHECK_FLAG_VALUE(EA_INODE);
479         CHECK_FLAG_VALUE(EOFBLOCKS);
480         CHECK_FLAG_VALUE(INLINE_DATA);
481         CHECK_FLAG_VALUE(RESERVED);
482 }
483
484 /* Used to pass group descriptor data when online resize is done */
485 struct ext4_new_group_input {
486         __u32 group;            /* Group number for this data */
487         __u64 block_bitmap;     /* Absolute block number of block bitmap */
488         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
489         __u64 inode_table;      /* Absolute block number of inode table start */
490         __u32 blocks_count;     /* Total number of blocks in this group */
491         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
492         __u16 unused;
493 };
494
495 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
496 struct compat_ext4_new_group_input {
497         u32 group;
498         compat_u64 block_bitmap;
499         compat_u64 inode_bitmap;
500         compat_u64 inode_table;
501         u32 blocks_count;
502         u16 reserved_blocks;
503         u16 unused;
504 };
505 #endif
506
507 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
508 struct ext4_new_group_data {
509         __u32 group;
510         __u64 block_bitmap;
511         __u64 inode_bitmap;
512         __u64 inode_table;
513         __u32 blocks_count;
514         __u16 reserved_blocks;
515         __u16 unused;
516         __u32 free_blocks_count;
517 };
518
519 /* Indexes used to index group tables in ext4_new_group_data */
520 enum {
521         BLOCK_BITMAP = 0,       /* block bitmap */
522         INODE_BITMAP,           /* inode bitmap */
523         INODE_TABLE,            /* inode tables */
524         GROUP_TABLE_COUNT,
525 };
526
527 /*
528  * Flags used by ext4_map_blocks()
529  */
530         /* Allocate any needed blocks and/or convert an unitialized
531            extent to be an initialized ext4 */
532 #define EXT4_GET_BLOCKS_CREATE                  0x0001
533         /* Request the creation of an unitialized extent */
534 #define EXT4_GET_BLOCKS_UNINIT_EXT              0x0002
535 #define EXT4_GET_BLOCKS_CREATE_UNINIT_EXT       (EXT4_GET_BLOCKS_UNINIT_EXT|\
536                                                  EXT4_GET_BLOCKS_CREATE)
537         /* Caller is from the delayed allocation writeout path
538          * finally doing the actual allocation of delayed blocks */
539 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
540         /* caller is from the direct IO path, request to creation of an
541         unitialized extents if not allocated, split the uninitialized
542         extent if blocks has been preallocated already*/
543 #define EXT4_GET_BLOCKS_PRE_IO                  0x0008
544 #define EXT4_GET_BLOCKS_CONVERT                 0x0010
545 #define EXT4_GET_BLOCKS_IO_CREATE_EXT           (EXT4_GET_BLOCKS_PRE_IO|\
546                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
547         /* Convert extent to initialized after IO complete */
548 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT          (EXT4_GET_BLOCKS_CONVERT|\
549                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
550         /* Eventual metadata allocation (due to growing extent tree)
551          * should not fail, so try to use reserved blocks for that.*/
552 #define EXT4_GET_BLOCKS_METADATA_NOFAIL         0x0020
553         /* Don't normalize allocation size (used for fallocate) */
554 #define EXT4_GET_BLOCKS_NO_NORMALIZE            0x0040
555         /* Request will not result in inode size update (user for fallocate) */
556 #define EXT4_GET_BLOCKS_KEEP_SIZE               0x0080
557         /* Do not take i_data_sem locking in ext4_map_blocks */
558 #define EXT4_GET_BLOCKS_NO_LOCK                 0x0100
559         /* Do not put hole in extent cache */
560 #define EXT4_GET_BLOCKS_NO_PUT_HOLE             0x0200
561
562 /*
563  * Flags used by ext4_free_blocks
564  */
565 #define EXT4_FREE_BLOCKS_METADATA       0x0001
566 #define EXT4_FREE_BLOCKS_FORGET         0x0002
567 #define EXT4_FREE_BLOCKS_VALIDATED      0x0004
568 #define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE 0x0008
569 #define EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER   0x0010
570 #define EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER    0x0020
571
572 /*
573  * ioctl commands
574  */
575 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
576 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
577 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
578 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
579 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
580 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
581 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
582 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
583 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
584 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
585 #define EXT4_IOC_MIGRATE                _IO('f', 9)
586  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
587  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
588 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
589 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
590 #define EXT4_IOC_RESIZE_FS              _IOW('f', 16, __u64)
591 #define EXT4_IOC_SWAP_BOOT              _IO('f', 17)
592
593 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
594 /*
595  * ioctl commands in 32 bit emulation
596  */
597 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
598 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
599 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
600 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
601 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
602 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
603 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
604 #define EXT4_IOC32_GROUP_ADD            _IOW('f', 8, struct compat_ext4_new_group_input)
605 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
606 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
607 #endif
608
609 /* Max physical block we can address w/o extents */
610 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
611
612 /*
613  * Structure of an inode on the disk
614  */
615 struct ext4_inode {
616         __le16  i_mode;         /* File mode */
617         __le16  i_uid;          /* Low 16 bits of Owner Uid */
618         __le32  i_size_lo;      /* Size in bytes */
619         __le32  i_atime;        /* Access time */
620         __le32  i_ctime;        /* Inode Change time */
621         __le32  i_mtime;        /* Modification time */
622         __le32  i_dtime;        /* Deletion Time */
623         __le16  i_gid;          /* Low 16 bits of Group Id */
624         __le16  i_links_count;  /* Links count */
625         __le32  i_blocks_lo;    /* Blocks count */
626         __le32  i_flags;        /* File flags */
627         union {
628                 struct {
629                         __le32  l_i_version;
630                 } linux1;
631                 struct {
632                         __u32  h_i_translator;
633                 } hurd1;
634                 struct {
635                         __u32  m_i_reserved1;
636                 } masix1;
637         } osd1;                         /* OS dependent 1 */
638         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
639         __le32  i_generation;   /* File version (for NFS) */
640         __le32  i_file_acl_lo;  /* File ACL */
641         __le32  i_size_high;
642         __le32  i_obso_faddr;   /* Obsoleted fragment address */
643         union {
644                 struct {
645                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
646                         __le16  l_i_file_acl_high;
647                         __le16  l_i_uid_high;   /* these 2 fields */
648                         __le16  l_i_gid_high;   /* were reserved2[0] */
649                         __le16  l_i_checksum_lo;/* crc32c(uuid+inum+inode) LE */
650                         __le16  l_i_reserved;
651                 } linux2;
652                 struct {
653                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
654                         __u16   h_i_mode_high;
655                         __u16   h_i_uid_high;
656                         __u16   h_i_gid_high;
657                         __u32   h_i_author;
658                 } hurd2;
659                 struct {
660                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
661                         __le16  m_i_file_acl_high;
662                         __u32   m_i_reserved2[2];
663                 } masix2;
664         } osd2;                         /* OS dependent 2 */
665         __le16  i_extra_isize;
666         __le16  i_checksum_hi;  /* crc32c(uuid+inum+inode) BE */
667         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
668         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
669         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
670         __le32  i_crtime;       /* File Creation time */
671         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
672         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
673 };
674
675 struct move_extent {
676         __u32 reserved;         /* should be zero */
677         __u32 donor_fd;         /* donor file descriptor */
678         __u64 orig_start;       /* logical start offset in block for orig */
679         __u64 donor_start;      /* logical start offset in block for donor */
680         __u64 len;              /* block length to be moved */
681         __u64 moved_len;        /* moved block length */
682 };
683
684 #define EXT4_EPOCH_BITS 2
685 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
686 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
687
688 /*
689  * Extended fields will fit into an inode if the filesystem was formatted
690  * with large inodes (-I 256 or larger) and there are not currently any EAs
691  * consuming all of the available space. For new inodes we always reserve
692  * enough space for the kernel's known extended fields, but for inodes
693  * created with an old kernel this might not have been the case. None of
694  * the extended inode fields is critical for correct filesystem operation.
695  * This macro checks if a certain field fits in the inode. Note that
696  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
697  */
698 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
699         ((offsetof(typeof(*ext4_inode), field) +        \
700           sizeof((ext4_inode)->field))                  \
701         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
702             (einode)->i_extra_isize))                   \
703
704 static inline __le32 ext4_encode_extra_time(struct timespec *time)
705 {
706        return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
707                            (time->tv_sec >> 32) & EXT4_EPOCH_MASK : 0) |
708                           ((time->tv_nsec << EXT4_EPOCH_BITS) & EXT4_NSEC_MASK));
709 }
710
711 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
712 {
713        if (sizeof(time->tv_sec) > 4)
714                time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
715                                << 32;
716        time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
717 }
718
719 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
720 do {                                                                           \
721         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
722         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
723                 (raw_inode)->xtime ## _extra =                                 \
724                                 ext4_encode_extra_time(&(inode)->xtime);       \
725 } while (0)
726
727 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
728 do {                                                                           \
729         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
730                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
731         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
732                 (raw_inode)->xtime ## _extra =                                 \
733                                 ext4_encode_extra_time(&(einode)->xtime);      \
734 } while (0)
735
736 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
737 do {                                                                           \
738         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
739         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
740                 ext4_decode_extra_time(&(inode)->xtime,                        \
741                                        raw_inode->xtime ## _extra);            \
742         else                                                                   \
743                 (inode)->xtime.tv_nsec = 0;                                    \
744 } while (0)
745
746 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
747 do {                                                                           \
748         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
749                 (einode)->xtime.tv_sec =                                       \
750                         (signed)le32_to_cpu((raw_inode)->xtime);               \
751         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
752                 ext4_decode_extra_time(&(einode)->xtime,                       \
753                                        raw_inode->xtime ## _extra);            \
754         else                                                                   \
755                 (einode)->xtime.tv_nsec = 0;                                   \
756 } while (0)
757
758 #define i_disk_version osd1.linux1.l_i_version
759
760 #if defined(__KERNEL__) || defined(__linux__)
761 #define i_reserved1     osd1.linux1.l_i_reserved1
762 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
763 #define i_blocks_high   osd2.linux2.l_i_blocks_high
764 #define i_uid_low       i_uid
765 #define i_gid_low       i_gid
766 #define i_uid_high      osd2.linux2.l_i_uid_high
767 #define i_gid_high      osd2.linux2.l_i_gid_high
768 #define i_checksum_lo   osd2.linux2.l_i_checksum_lo
769
770 #elif defined(__GNU__)
771
772 #define i_translator    osd1.hurd1.h_i_translator
773 #define i_uid_high      osd2.hurd2.h_i_uid_high
774 #define i_gid_high      osd2.hurd2.h_i_gid_high
775 #define i_author        osd2.hurd2.h_i_author
776
777 #elif defined(__masix__)
778
779 #define i_reserved1     osd1.masix1.m_i_reserved1
780 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
781 #define i_reserved2     osd2.masix2.m_i_reserved2
782
783 #endif /* defined(__KERNEL__) || defined(__linux__) */
784
785 #include "extents_status.h"
786
787 /*
788  * fourth extended file system inode data in memory
789  */
790 struct ext4_inode_info {
791         __le32  i_data[15];     /* unconverted */
792         __u32   i_dtime;
793         ext4_fsblk_t    i_file_acl;
794
795         /*
796          * i_block_group is the number of the block group which contains
797          * this file's inode.  Constant across the lifetime of the inode,
798          * it is ued for making block allocation decisions - we try to
799          * place a file's data blocks near its inode block, and new inodes
800          * near to their parent directory's inode.
801          */
802         ext4_group_t    i_block_group;
803         ext4_lblk_t     i_dir_start_lookup;
804 #if (BITS_PER_LONG < 64)
805         unsigned long   i_state_flags;          /* Dynamic state flags */
806 #endif
807         unsigned long   i_flags;
808
809         /*
810          * Extended attributes can be read independently of the main file
811          * data. Taking i_mutex even when reading would cause contention
812          * between readers of EAs and writers of regular file data, so
813          * instead we synchronize on xattr_sem when reading or changing
814          * EAs.
815          */
816         struct rw_semaphore xattr_sem;
817
818         struct list_head i_orphan;      /* unlinked but open inodes */
819
820         /*
821          * i_disksize keeps track of what the inode size is ON DISK, not
822          * in memory.  During truncate, i_size is set to the new size by
823          * the VFS prior to calling ext4_truncate(), but the filesystem won't
824          * set i_disksize to 0 until the truncate is actually under way.
825          *
826          * The intent is that i_disksize always represents the blocks which
827          * are used by this file.  This allows recovery to restart truncate
828          * on orphans if we crash during truncate.  We actually write i_disksize
829          * into the on-disk inode when writing inodes out, instead of i_size.
830          *
831          * The only time when i_disksize and i_size may be different is when
832          * a truncate is in progress.  The only things which change i_disksize
833          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
834          */
835         loff_t  i_disksize;
836
837         /*
838          * i_data_sem is for serialising ext4_truncate() against
839          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
840          * data tree are chopped off during truncate. We can't do that in
841          * ext4 because whenever we perform intermediate commits during
842          * truncate, the inode and all the metadata blocks *must* be in a
843          * consistent state which allows truncation of the orphans to restart
844          * during recovery.  Hence we must fix the get_block-vs-truncate race
845          * by other means, so we have i_data_sem.
846          */
847         struct rw_semaphore i_data_sem;
848         struct inode vfs_inode;
849         struct jbd2_inode *jinode;
850
851         /*
852          * File creation time. Its function is same as that of
853          * struct timespec i_{a,c,m}time in the generic inode.
854          */
855         struct timespec i_crtime;
856
857         /* mballoc */
858         struct list_head i_prealloc_list;
859         spinlock_t i_prealloc_lock;
860
861         /* extents status tree */
862         struct ext4_es_tree i_es_tree;
863         rwlock_t i_es_lock;
864         struct list_head i_es_lru;
865         unsigned int i_es_lru_nr;       /* protected by i_es_lock */
866
867         /* ialloc */
868         ext4_group_t    i_last_alloc_group;
869
870         /* allocation reservation info for delalloc */
871         /* In case of bigalloc, these refer to clusters rather than blocks */
872         unsigned int i_reserved_data_blocks;
873         unsigned int i_reserved_meta_blocks;
874         unsigned int i_allocated_meta_blocks;
875         ext4_lblk_t i_da_metadata_calc_last_lblock;
876         int i_da_metadata_calc_len;
877
878         /* on-disk additional length */
879         __u16 i_extra_isize;
880
881         /* Indicate the inline data space. */
882         u16 i_inline_off;
883         u16 i_inline_size;
884
885 #ifdef CONFIG_QUOTA
886         /* quota space reservation, managed internally by quota code */
887         qsize_t i_reserved_quota;
888 #endif
889
890         /* Lock protecting lists below */
891         spinlock_t i_completed_io_lock;
892         /*
893          * Completed IOs that need unwritten extents handling and have
894          * transaction reserved
895          */
896         struct list_head i_rsv_conversion_list;
897         /*
898          * Completed IOs that need unwritten extents handling and don't have
899          * transaction reserved
900          */
901         struct list_head i_unrsv_conversion_list;
902         atomic_t i_ioend_count; /* Number of outstanding io_end structs */
903         atomic_t i_unwritten; /* Nr. of inflight conversions pending */
904         struct work_struct i_rsv_conversion_work;
905         struct work_struct i_unrsv_conversion_work;
906
907         spinlock_t i_block_reservation_lock;
908
909         /*
910          * Transactions that contain inode's metadata needed to complete
911          * fsync and fdatasync, respectively.
912          */
913         tid_t i_sync_tid;
914         tid_t i_datasync_tid;
915
916         /* Precomputed uuid+inum+igen checksum for seeding inode checksums */
917         __u32 i_csum_seed;
918 };
919
920 /*
921  * File system states
922  */
923 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
924 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
925 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
926
927 /*
928  * Misc. filesystem flags
929  */
930 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
931 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
932 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
933
934 /*
935  * Mount flags set via mount options or defaults
936  */
937 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
938 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
939 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
940 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
941 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
942 #define EXT4_MOUNT_ERRORS_MASK          0x00070
943 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
944 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
945 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
946 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
947 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
948 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
949 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
950 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
951 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
952 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
953 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
954 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
955 #define EXT4_MOUNT_QUOTA                0x80000 /* Some quota option set */
956 #define EXT4_MOUNT_USRQUOTA             0x100000 /* "old" user quota */
957 #define EXT4_MOUNT_GRPQUOTA             0x200000 /* "old" group quota */
958 #define EXT4_MOUNT_DIOREAD_NOLOCK       0x400000 /* Enable support for dio read nolocking */
959 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
960 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
961 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
962 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
963 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
964 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
965 #define EXT4_MOUNT_INIT_INODE_TABLE     0x80000000 /* Initialize uninitialized itables */
966
967 /*
968  * Mount flags set either automatically (could not be set by mount option)
969  * based on per file system feature or property or in special cases such as
970  * distinguishing between explicit mount option definition and default.
971  */
972 #define EXT4_MOUNT2_EXPLICIT_DELALLOC   0x00000001 /* User explicitly
973                                                       specified delalloc */
974 #define EXT4_MOUNT2_STD_GROUP_SIZE      0x00000002 /* We have standard group
975                                                       size of blocksize * 8
976                                                       blocks */
977
978 #define clear_opt(sb, opt)              EXT4_SB(sb)->s_mount_opt &= \
979                                                 ~EXT4_MOUNT_##opt
980 #define set_opt(sb, opt)                EXT4_SB(sb)->s_mount_opt |= \
981                                                 EXT4_MOUNT_##opt
982 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
983                                          EXT4_MOUNT_##opt)
984
985 #define clear_opt2(sb, opt)             EXT4_SB(sb)->s_mount_opt2 &= \
986                                                 ~EXT4_MOUNT2_##opt
987 #define set_opt2(sb, opt)               EXT4_SB(sb)->s_mount_opt2 |= \
988                                                 EXT4_MOUNT2_##opt
989 #define test_opt2(sb, opt)              (EXT4_SB(sb)->s_mount_opt2 & \
990                                          EXT4_MOUNT2_##opt)
991
992 #define ext4_test_and_set_bit           __test_and_set_bit_le
993 #define ext4_set_bit                    __set_bit_le
994 #define ext4_set_bit_atomic             ext2_set_bit_atomic
995 #define ext4_test_and_clear_bit         __test_and_clear_bit_le
996 #define ext4_clear_bit                  __clear_bit_le
997 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
998 #define ext4_test_bit                   test_bit_le
999 #define ext4_find_next_zero_bit         find_next_zero_bit_le
1000 #define ext4_find_next_bit              find_next_bit_le
1001
1002 extern void ext4_set_bits(void *bm, int cur, int len);
1003
1004 /*
1005  * Maximal mount counts between two filesystem checks
1006  */
1007 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
1008 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
1009
1010 /*
1011  * Behaviour when detecting errors
1012  */
1013 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
1014 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
1015 #define EXT4_ERRORS_PANIC               3       /* Panic */
1016 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
1017
1018 /* Metadata checksum algorithm codes */
1019 #define EXT4_CRC32C_CHKSUM              1
1020
1021 /*
1022  * Structure of the super block
1023  */
1024 struct ext4_super_block {
1025 /*00*/  __le32  s_inodes_count;         /* Inodes count */
1026         __le32  s_blocks_count_lo;      /* Blocks count */
1027         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
1028         __le32  s_free_blocks_count_lo; /* Free blocks count */
1029 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
1030         __le32  s_first_data_block;     /* First Data Block */
1031         __le32  s_log_block_size;       /* Block size */
1032         __le32  s_log_cluster_size;     /* Allocation cluster size */
1033 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
1034         __le32  s_clusters_per_group;   /* # Clusters per group */
1035         __le32  s_inodes_per_group;     /* # Inodes per group */
1036         __le32  s_mtime;                /* Mount time */
1037 /*30*/  __le32  s_wtime;                /* Write time */
1038         __le16  s_mnt_count;            /* Mount count */
1039         __le16  s_max_mnt_count;        /* Maximal mount count */
1040         __le16  s_magic;                /* Magic signature */
1041         __le16  s_state;                /* File system state */
1042         __le16  s_errors;               /* Behaviour when detecting errors */
1043         __le16  s_minor_rev_level;      /* minor revision level */
1044 /*40*/  __le32  s_lastcheck;            /* time of last check */
1045         __le32  s_checkinterval;        /* max. time between checks */
1046         __le32  s_creator_os;           /* OS */
1047         __le32  s_rev_level;            /* Revision level */
1048 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
1049         __le16  s_def_resgid;           /* Default gid for reserved blocks */
1050         /*
1051          * These fields are for EXT4_DYNAMIC_REV superblocks only.
1052          *
1053          * Note: the difference between the compatible feature set and
1054          * the incompatible feature set is that if there is a bit set
1055          * in the incompatible feature set that the kernel doesn't
1056          * know about, it should refuse to mount the filesystem.
1057          *
1058          * e2fsck's requirements are more strict; if it doesn't know
1059          * about a feature in either the compatible or incompatible
1060          * feature set, it must abort and not try to meddle with
1061          * things it doesn't understand...
1062          */
1063         __le32  s_first_ino;            /* First non-reserved inode */
1064         __le16  s_inode_size;           /* size of inode structure */
1065         __le16  s_block_group_nr;       /* block group # of this superblock */
1066         __le32  s_feature_compat;       /* compatible feature set */
1067 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
1068         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
1069 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
1070 /*78*/  char    s_volume_name[16];      /* volume name */
1071 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
1072 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
1073         /*
1074          * Performance hints.  Directory preallocation should only
1075          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1076          */
1077         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
1078         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
1079         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
1080         /*
1081          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1082          */
1083 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
1084 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
1085         __le32  s_journal_dev;          /* device number of journal file */
1086         __le32  s_last_orphan;          /* start of list of inodes to delete */
1087         __le32  s_hash_seed[4];         /* HTREE hash seed */
1088         __u8    s_def_hash_version;     /* Default hash version to use */
1089         __u8    s_jnl_backup_type;
1090         __le16  s_desc_size;            /* size of group descriptor */
1091 /*100*/ __le32  s_default_mount_opts;
1092         __le32  s_first_meta_bg;        /* First metablock block group */
1093         __le32  s_mkfs_time;            /* When the filesystem was created */
1094         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
1095         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
1096 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
1097         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
1098         __le32  s_free_blocks_count_hi; /* Free blocks count */
1099         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
1100         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
1101         __le32  s_flags;                /* Miscellaneous flags */
1102         __le16  s_raid_stride;          /* RAID stride */
1103         __le16  s_mmp_update_interval;  /* # seconds to wait in MMP checking */
1104         __le64  s_mmp_block;            /* Block for multi-mount protection */
1105         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
1106         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
1107         __u8    s_checksum_type;        /* metadata checksum algorithm used */
1108         __le16  s_reserved_pad;
1109         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
1110         __le32  s_snapshot_inum;        /* Inode number of active snapshot */
1111         __le32  s_snapshot_id;          /* sequential ID of active snapshot */
1112         __le64  s_snapshot_r_blocks_count; /* reserved blocks for active
1113                                               snapshot's future use */
1114         __le32  s_snapshot_list;        /* inode number of the head of the
1115                                            on-disk snapshot list */
1116 #define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1117         __le32  s_error_count;          /* number of fs errors */
1118         __le32  s_first_error_time;     /* first time an error happened */
1119         __le32  s_first_error_ino;      /* inode involved in first error */
1120         __le64  s_first_error_block;    /* block involved of first error */
1121         __u8    s_first_error_func[32]; /* function where the error happened */
1122         __le32  s_first_error_line;     /* line number where error happened */
1123         __le32  s_last_error_time;      /* most recent time of an error */
1124         __le32  s_last_error_ino;       /* inode involved in last error */
1125         __le32  s_last_error_line;      /* line number where error happened */
1126         __le64  s_last_error_block;     /* block involved of last error */
1127         __u8    s_last_error_func[32];  /* function where the error happened */
1128 #define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1129         __u8    s_mount_opts[64];
1130         __le32  s_usr_quota_inum;       /* inode for tracking user quota */
1131         __le32  s_grp_quota_inum;       /* inode for tracking group quota */
1132         __le32  s_overhead_clusters;    /* overhead blocks/clusters in fs */
1133         __le32  s_reserved[108];        /* Padding to the end of the block */
1134         __le32  s_checksum;             /* crc32c(superblock) */
1135 };
1136
1137 #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1138
1139 #ifdef __KERNEL__
1140
1141 /*
1142  * run-time mount flags
1143  */
1144 #define EXT4_MF_MNTDIR_SAMPLED  0x0001
1145 #define EXT4_MF_FS_ABORTED      0x0002  /* Fatal error detected */
1146
1147 /*
1148  * fourth extended-fs super-block data in memory
1149  */
1150 struct ext4_sb_info {
1151         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
1152         unsigned long s_inodes_per_block;/* Number of inodes per block */
1153         unsigned long s_blocks_per_group;/* Number of blocks in a group */
1154         unsigned long s_clusters_per_group; /* Number of clusters in a group */
1155         unsigned long s_inodes_per_group;/* Number of inodes in a group */
1156         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
1157         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
1158         unsigned long s_desc_per_block; /* Number of group descriptors per block */
1159         ext4_group_t s_groups_count;    /* Number of groups in the fs */
1160         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1161         unsigned long s_overhead;  /* # of fs overhead clusters */
1162         unsigned int s_cluster_ratio;   /* Number of blocks per cluster */
1163         unsigned int s_cluster_bits;    /* log2 of s_cluster_ratio */
1164         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
1165         struct buffer_head * s_sbh;     /* Buffer containing the super block */
1166         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
1167         struct buffer_head **s_group_desc;
1168         unsigned int s_mount_opt;
1169         unsigned int s_mount_opt2;
1170         unsigned int s_mount_flags;
1171         unsigned int s_def_mount_opt;
1172         ext4_fsblk_t s_sb_block;
1173         atomic64_t s_resv_clusters;
1174         kuid_t s_resuid;
1175         kgid_t s_resgid;
1176         unsigned short s_mount_state;
1177         unsigned short s_pad;
1178         int s_addr_per_block_bits;
1179         int s_desc_per_block_bits;
1180         int s_inode_size;
1181         int s_first_ino;
1182         unsigned int s_inode_readahead_blks;
1183         unsigned int s_inode_goal;
1184         spinlock_t s_next_gen_lock;
1185         u32 s_next_generation;
1186         u32 s_hash_seed[4];
1187         int s_def_hash_version;
1188         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
1189         struct percpu_counter s_freeclusters_counter;
1190         struct percpu_counter s_freeinodes_counter;
1191         struct percpu_counter s_dirs_counter;
1192         struct percpu_counter s_dirtyclusters_counter;
1193         struct blockgroup_lock *s_blockgroup_lock;
1194         struct proc_dir_entry *s_proc;
1195         struct kobject s_kobj;
1196         struct completion s_kobj_unregister;
1197         struct super_block *s_sb;
1198
1199         /* Journaling */
1200         struct journal_s *s_journal;
1201         struct list_head s_orphan;
1202         struct mutex s_orphan_lock;
1203         unsigned long s_resize_flags;           /* Flags indicating if there
1204                                                    is a resizer */
1205         unsigned long s_commit_interval;
1206         u32 s_max_batch_time;
1207         u32 s_min_batch_time;
1208         struct block_device *journal_bdev;
1209 #ifdef CONFIG_QUOTA
1210         char *s_qf_names[MAXQUOTAS];            /* Names of quota files with journalled quota */
1211         int s_jquota_fmt;                       /* Format of quota to use */
1212 #endif
1213         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1214         struct rb_root system_blks;
1215
1216 #ifdef EXTENTS_STATS
1217         /* ext4 extents stats */
1218         unsigned long s_ext_min;
1219         unsigned long s_ext_max;
1220         unsigned long s_depth_max;
1221         spinlock_t s_ext_stats_lock;
1222         unsigned long s_ext_blocks;
1223         unsigned long s_ext_extents;
1224 #endif
1225
1226         /* for buddy allocator */
1227         struct ext4_group_info ***s_group_info;
1228         struct inode *s_buddy_cache;
1229         spinlock_t s_md_lock;
1230         unsigned short *s_mb_offsets;
1231         unsigned int *s_mb_maxs;
1232         unsigned int s_group_info_size;
1233
1234         /* tunables */
1235         unsigned long s_stripe;
1236         unsigned int s_mb_stream_request;
1237         unsigned int s_mb_max_to_scan;
1238         unsigned int s_mb_min_to_scan;
1239         unsigned int s_mb_stats;
1240         unsigned int s_mb_order2_reqs;
1241         unsigned int s_mb_group_prealloc;
1242         unsigned int s_max_dir_size_kb;
1243         /* where last allocation was done - for stream allocation */
1244         unsigned long s_mb_last_group;
1245         unsigned long s_mb_last_start;
1246
1247         /* stats for buddy allocator */
1248         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
1249         atomic_t s_bal_success; /* we found long enough chunks */
1250         atomic_t s_bal_allocated;       /* in blocks */
1251         atomic_t s_bal_ex_scanned;      /* total extents scanned */
1252         atomic_t s_bal_goals;   /* goal hits */
1253         atomic_t s_bal_breaks;  /* too long searches */
1254         atomic_t s_bal_2orders; /* 2^order hits */
1255         spinlock_t s_bal_lock;
1256         unsigned long s_mb_buddies_generated;
1257         unsigned long long s_mb_generation_time;
1258         atomic_t s_mb_lost_chunks;
1259         atomic_t s_mb_preallocated;
1260         atomic_t s_mb_discarded;
1261         atomic_t s_lock_busy;
1262
1263         /* locality groups */
1264         struct ext4_locality_group __percpu *s_locality_groups;
1265
1266         /* for write statistics */
1267         unsigned long s_sectors_written_start;
1268         u64 s_kbytes_written;
1269
1270         /* the size of zero-out chunk */
1271         unsigned int s_extent_max_zeroout_kb;
1272
1273         unsigned int s_log_groups_per_flex;
1274         struct flex_groups *s_flex_groups;
1275         ext4_group_t s_flex_groups_allocated;
1276
1277         /* workqueue for unreserved extent convertions (dio) */
1278         struct workqueue_struct *unrsv_conversion_wq;
1279         /* workqueue for reserved extent conversions (buffered io) */
1280         struct workqueue_struct *rsv_conversion_wq;
1281
1282         /* timer for periodic error stats printing */
1283         struct timer_list s_err_report;
1284
1285         /* Lazy inode table initialization info */
1286         struct ext4_li_request *s_li_request;
1287         /* Wait multiplier for lazy initialization thread */
1288         unsigned int s_li_wait_mult;
1289
1290         /* Kernel thread for multiple mount protection */
1291         struct task_struct *s_mmp_tsk;
1292
1293         /* record the last minlen when FITRIM is called. */
1294         atomic_t s_last_trim_minblks;
1295
1296         /* Reference to checksum algorithm driver via cryptoapi */
1297         struct crypto_shash *s_chksum_driver;
1298
1299         /* Precomputed FS UUID checksum for seeding other checksums */
1300         __u32 s_csum_seed;
1301
1302         /* Reclaim extents from extent status tree */
1303         struct shrinker s_es_shrinker;
1304         struct list_head s_es_lru;
1305         struct percpu_counter s_extent_cache_cnt;
1306         spinlock_t s_es_lru_lock ____cacheline_aligned_in_smp;
1307 };
1308
1309 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1310 {
1311         return sb->s_fs_info;
1312 }
1313 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1314 {
1315         return container_of(inode, struct ext4_inode_info, vfs_inode);
1316 }
1317
1318 static inline struct timespec ext4_current_time(struct inode *inode)
1319 {
1320         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1321                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1322 }
1323
1324 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1325 {
1326         return ino == EXT4_ROOT_INO ||
1327                 ino == EXT4_USR_QUOTA_INO ||
1328                 ino == EXT4_GRP_QUOTA_INO ||
1329                 ino == EXT4_BOOT_LOADER_INO ||
1330                 ino == EXT4_JOURNAL_INO ||
1331                 ino == EXT4_RESIZE_INO ||
1332                 (ino >= EXT4_FIRST_INO(sb) &&
1333                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1334 }
1335
1336 static inline void ext4_set_io_unwritten_flag(struct inode *inode,
1337                                               struct ext4_io_end *io_end)
1338 {
1339         if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
1340                 /* Writeback has to have coversion transaction reserved */
1341                 WARN_ON(EXT4_SB(inode->i_sb)->s_journal && !io_end->handle &&
1342                         !(io_end->flag & EXT4_IO_END_DIRECT));
1343                 io_end->flag |= EXT4_IO_END_UNWRITTEN;
1344                 atomic_inc(&EXT4_I(inode)->i_unwritten);
1345         }
1346 }
1347
1348 static inline ext4_io_end_t *ext4_inode_aio(struct inode *inode)
1349 {
1350         return inode->i_private;
1351 }
1352
1353 static inline void ext4_inode_aio_set(struct inode *inode, ext4_io_end_t *io)
1354 {
1355         inode->i_private = io;
1356 }
1357
1358 /*
1359  * Inode dynamic state flags
1360  */
1361 enum {
1362         EXT4_STATE_JDATA,               /* journaled data exists */
1363         EXT4_STATE_NEW,                 /* inode is newly created */
1364         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1365         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1366         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1367         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1368         EXT4_STATE_DIO_UNWRITTEN,       /* need convert on dio done*/
1369         EXT4_STATE_NEWENTRY,            /* File just added to dir */
1370         EXT4_STATE_DELALLOC_RESERVED,   /* blks already reserved for delalloc */
1371         EXT4_STATE_DIOREAD_LOCK,        /* Disable support for dio read
1372                                            nolocking */
1373         EXT4_STATE_MAY_INLINE_DATA,     /* may have in-inode data */
1374         EXT4_STATE_ORDERED_MODE,        /* data=ordered mode */
1375 };
1376
1377 #define EXT4_INODE_BIT_FNS(name, field, offset)                         \
1378 static inline int ext4_test_inode_##name(struct inode *inode, int bit)  \
1379 {                                                                       \
1380         return test_bit(bit + (offset), &EXT4_I(inode)->i_##field);     \
1381 }                                                                       \
1382 static inline void ext4_set_inode_##name(struct inode *inode, int bit)  \
1383 {                                                                       \
1384         set_bit(bit + (offset), &EXT4_I(inode)->i_##field);             \
1385 }                                                                       \
1386 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1387 {                                                                       \
1388         clear_bit(bit + (offset), &EXT4_I(inode)->i_##field);           \
1389 }
1390
1391 EXT4_INODE_BIT_FNS(flag, flags, 0)
1392 #if (BITS_PER_LONG < 64)
1393 EXT4_INODE_BIT_FNS(state, state_flags, 0)
1394
1395 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1396 {
1397         (ei)->i_state_flags = 0;
1398 }
1399 #else
1400 EXT4_INODE_BIT_FNS(state, flags, 32)
1401
1402 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1403 {
1404         /* We depend on the fact that callers will set i_flags */
1405 }
1406 #endif
1407 #else
1408 /* Assume that user mode programs are passing in an ext4fs superblock, not
1409  * a kernel struct super_block.  This will allow us to call the feature-test
1410  * macros from user land. */
1411 #define EXT4_SB(sb)     (sb)
1412 #endif
1413
1414 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1415
1416 /*
1417  * Codes for operating systems
1418  */
1419 #define EXT4_OS_LINUX           0
1420 #define EXT4_OS_HURD            1
1421 #define EXT4_OS_MASIX           2
1422 #define EXT4_OS_FREEBSD         3
1423 #define EXT4_OS_LITES           4
1424
1425 /*
1426  * Revision levels
1427  */
1428 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1429 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1430
1431 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1432 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1433
1434 #define EXT4_GOOD_OLD_INODE_SIZE 128
1435
1436 /*
1437  * Feature set definitions
1438  */
1439
1440 #define EXT4_HAS_COMPAT_FEATURE(sb,mask)                        \
1441         ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1442 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask)                     \
1443         ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1444 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask)                      \
1445         ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1446 #define EXT4_SET_COMPAT_FEATURE(sb,mask)                        \
1447         EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1448 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask)                     \
1449         EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1450 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask)                      \
1451         EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1452 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask)                      \
1453         EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1454 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask)                   \
1455         EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1456 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask)                    \
1457         EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1458
1459 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1460 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1461 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1462 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1463 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1464 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1465
1466 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1467 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1468 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1469 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1470 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1471 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1472 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1473 #define EXT4_FEATURE_RO_COMPAT_QUOTA            0x0100
1474 #define EXT4_FEATURE_RO_COMPAT_BIGALLOC         0x0200
1475 /*
1476  * METADATA_CSUM also enables group descriptor checksums (GDT_CSUM).  When
1477  * METADATA_CSUM is set, group descriptor checksums use the same algorithm as
1478  * all other data structures' checksums.  However, the METADATA_CSUM and
1479  * GDT_CSUM bits are mutually exclusive.
1480  */
1481 #define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM    0x0400
1482
1483 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1484 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1485 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1486 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1487 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1488 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1489 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1490 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1491 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1492 #define EXT4_FEATURE_INCOMPAT_EA_INODE          0x0400 /* EA in inode */
1493 #define EXT4_FEATURE_INCOMPAT_DIRDATA           0x1000 /* data in dirent */
1494 #define EXT4_FEATURE_INCOMPAT_BG_USE_META_CSUM  0x2000 /* use crc32c for bg */
1495 #define EXT4_FEATURE_INCOMPAT_LARGEDIR          0x4000 /* >2GB or 3-lvl htree */
1496 #define EXT4_FEATURE_INCOMPAT_INLINE_DATA       0x8000 /* data in inode */
1497
1498 #define EXT2_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1499 #define EXT2_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1500                                          EXT4_FEATURE_INCOMPAT_META_BG)
1501 #define EXT2_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1502                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1503                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1504
1505 #define EXT3_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1506 #define EXT3_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1507                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1508                                          EXT4_FEATURE_INCOMPAT_META_BG)
1509 #define EXT3_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1510                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1511                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1512
1513 #define EXT4_FEATURE_COMPAT_SUPP        EXT2_FEATURE_COMPAT_EXT_ATTR
1514 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1515                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1516                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1517                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1518                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1519                                          EXT4_FEATURE_INCOMPAT_FLEX_BG| \
1520                                          EXT4_FEATURE_INCOMPAT_MMP |    \
1521                                          EXT4_FEATURE_INCOMPAT_INLINE_DATA)
1522 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1523                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1524                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1525                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1526                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1527                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1528                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\
1529                                          EXT4_FEATURE_RO_COMPAT_BIGALLOC |\
1530                                          EXT4_FEATURE_RO_COMPAT_METADATA_CSUM|\
1531                                          EXT4_FEATURE_RO_COMPAT_QUOTA)
1532
1533 /*
1534  * Default values for user and/or group using reserved blocks
1535  */
1536 #define EXT4_DEF_RESUID         0
1537 #define EXT4_DEF_RESGID         0
1538
1539 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1540
1541 /*
1542  * Default mount options
1543  */
1544 #define EXT4_DEFM_DEBUG         0x0001
1545 #define EXT4_DEFM_BSDGROUPS     0x0002
1546 #define EXT4_DEFM_XATTR_USER    0x0004
1547 #define EXT4_DEFM_ACL           0x0008
1548 #define EXT4_DEFM_UID16         0x0010
1549 #define EXT4_DEFM_JMODE         0x0060
1550 #define EXT4_DEFM_JMODE_DATA    0x0020
1551 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1552 #define EXT4_DEFM_JMODE_WBACK   0x0060
1553 #define EXT4_DEFM_NOBARRIER     0x0100
1554 #define EXT4_DEFM_BLOCK_VALIDITY 0x0200
1555 #define EXT4_DEFM_DISCARD       0x0400
1556 #define EXT4_DEFM_NODELALLOC    0x0800
1557
1558 /*
1559  * Default journal batch times
1560  */
1561 #define EXT4_DEF_MIN_BATCH_TIME 0
1562 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1563
1564 /*
1565  * Minimum number of groups in a flexgroup before we separate out
1566  * directories into the first block group of a flexgroup
1567  */
1568 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1569
1570 /*
1571  * Structure of a directory entry
1572  */
1573 #define EXT4_NAME_LEN 255
1574
1575 struct ext4_dir_entry {
1576         __le32  inode;                  /* Inode number */
1577         __le16  rec_len;                /* Directory entry length */
1578         __le16  name_len;               /* Name length */
1579         char    name[EXT4_NAME_LEN];    /* File name */
1580 };
1581
1582 /*
1583  * The new version of the directory entry.  Since EXT4 structures are
1584  * stored in intel byte order, and the name_len field could never be
1585  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1586  * file_type field.
1587  */
1588 struct ext4_dir_entry_2 {
1589         __le32  inode;                  /* Inode number */
1590         __le16  rec_len;                /* Directory entry length */
1591         __u8    name_len;               /* Name length */
1592         __u8    file_type;
1593         char    name[EXT4_NAME_LEN];    /* File name */
1594 };
1595
1596 /*
1597  * This is a bogus directory entry at the end of each leaf block that
1598  * records checksums.
1599  */
1600 struct ext4_dir_entry_tail {
1601         __le32  det_reserved_zero1;     /* Pretend to be unused */
1602         __le16  det_rec_len;            /* 12 */
1603         __u8    det_reserved_zero2;     /* Zero name length */
1604         __u8    det_reserved_ft;        /* 0xDE, fake file type */
1605         __le32  det_checksum;           /* crc32c(uuid+inum+dirblock) */
1606 };
1607
1608 #define EXT4_DIRENT_TAIL(block, blocksize) \
1609         ((struct ext4_dir_entry_tail *)(((void *)(block)) + \
1610                                         ((blocksize) - \
1611                                          sizeof(struct ext4_dir_entry_tail))))
1612
1613 /*
1614  * Ext4 directory file types.  Only the low 3 bits are used.  The
1615  * other bits are reserved for now.
1616  */
1617 #define EXT4_FT_UNKNOWN         0
1618 #define EXT4_FT_REG_FILE        1
1619 #define EXT4_FT_DIR             2
1620 #define EXT4_FT_CHRDEV          3
1621 #define EXT4_FT_BLKDEV          4
1622 #define EXT4_FT_FIFO            5
1623 #define EXT4_FT_SOCK            6
1624 #define EXT4_FT_SYMLINK         7
1625
1626 #define EXT4_FT_MAX             8
1627
1628 #define EXT4_FT_DIR_CSUM        0xDE
1629
1630 /*
1631  * EXT4_DIR_PAD defines the directory entries boundaries
1632  *
1633  * NOTE: It must be a multiple of 4
1634  */
1635 #define EXT4_DIR_PAD                    4
1636 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1637 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1638                                          ~EXT4_DIR_ROUND)
1639 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1640
1641 /*
1642  * If we ever get support for fs block sizes > page_size, we'll need
1643  * to remove the #if statements in the next two functions...
1644  */
1645 static inline unsigned int
1646 ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
1647 {
1648         unsigned len = le16_to_cpu(dlen);
1649
1650 #if (PAGE_CACHE_SIZE >= 65536)
1651         if (len == EXT4_MAX_REC_LEN || len == 0)
1652                 return blocksize;
1653         return (len & 65532) | ((len & 3) << 16);
1654 #else
1655         return len;
1656 #endif
1657 }
1658
1659 static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
1660 {
1661         if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
1662                 BUG();
1663 #if (PAGE_CACHE_SIZE >= 65536)
1664         if (len < 65536)
1665                 return cpu_to_le16(len);
1666         if (len == blocksize) {
1667                 if (blocksize == 65536)
1668                         return cpu_to_le16(EXT4_MAX_REC_LEN);
1669                 else
1670                         return cpu_to_le16(0);
1671         }
1672         return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
1673 #else
1674         return cpu_to_le16(len);
1675 #endif
1676 }
1677
1678 /*
1679  * Hash Tree Directory indexing
1680  * (c) Daniel Phillips, 2001
1681  */
1682
1683 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1684                                       EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1685                     ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
1686 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1687 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1688
1689 /* Legal values for the dx_root hash_version field: */
1690
1691 #define DX_HASH_LEGACY          0
1692 #define DX_HASH_HALF_MD4        1
1693 #define DX_HASH_TEA             2
1694 #define DX_HASH_LEGACY_UNSIGNED 3
1695 #define DX_HASH_HALF_MD4_UNSIGNED       4
1696 #define DX_HASH_TEA_UNSIGNED            5
1697
1698 static inline u32 ext4_chksum(struct ext4_sb_info *sbi, u32 crc,
1699                               const void *address, unsigned int length)
1700 {
1701         struct {
1702                 struct shash_desc shash;
1703                 char ctx[4];
1704         } desc;
1705         int err;
1706
1707         BUG_ON(crypto_shash_descsize(sbi->s_chksum_driver)!=sizeof(desc.ctx));
1708
1709         desc.shash.tfm = sbi->s_chksum_driver;
1710         desc.shash.flags = 0;
1711         *(u32 *)desc.ctx = crc;
1712
1713         err = crypto_shash_update(&desc.shash, address, length);
1714         BUG_ON(err);
1715
1716         return *(u32 *)desc.ctx;
1717 }
1718
1719 #ifdef __KERNEL__
1720
1721 /* hash info structure used by the directory hash */
1722 struct dx_hash_info
1723 {
1724         u32             hash;
1725         u32             minor_hash;
1726         int             hash_version;
1727         u32             *seed;
1728 };
1729
1730
1731 /* 32 and 64 bit signed EOF for dx directories */
1732 #define EXT4_HTREE_EOF_32BIT   ((1UL  << (32 - 1)) - 1)
1733 #define EXT4_HTREE_EOF_64BIT   ((1ULL << (64 - 1)) - 1)
1734
1735
1736 /*
1737  * Control parameters used by ext4_htree_next_block
1738  */
1739 #define HASH_NB_ALWAYS          1
1740
1741
1742 /*
1743  * Describe an inode's exact location on disk and in memory
1744  */
1745 struct ext4_iloc
1746 {
1747         struct buffer_head *bh;
1748         unsigned long offset;
1749         ext4_group_t block_group;
1750 };
1751
1752 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1753 {
1754         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1755 }
1756
1757 /*
1758  * This structure is stuffed into the struct file's private_data field
1759  * for directories.  It is where we put information so that we can do
1760  * readdir operations in hash tree order.
1761  */
1762 struct dir_private_info {
1763         struct rb_root  root;
1764         struct rb_node  *curr_node;
1765         struct fname    *extra_fname;
1766         loff_t          last_pos;
1767         __u32           curr_hash;
1768         __u32           curr_minor_hash;
1769         __u32           next_hash;
1770 };
1771
1772 /* calculate the first block number of the group */
1773 static inline ext4_fsblk_t
1774 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1775 {
1776         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1777                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1778 }
1779
1780 /*
1781  * Special error return code only used by dx_probe() and its callers.
1782  */
1783 #define ERR_BAD_DX_DIR  -75000
1784
1785 /*
1786  * Timeout and state flag for lazy initialization inode thread.
1787  */
1788 #define EXT4_DEF_LI_WAIT_MULT                   10
1789 #define EXT4_DEF_LI_MAX_START_DELAY             5
1790 #define EXT4_LAZYINIT_QUIT                      0x0001
1791 #define EXT4_LAZYINIT_RUNNING                   0x0002
1792
1793 /*
1794  * Lazy inode table initialization info
1795  */
1796 struct ext4_lazy_init {
1797         unsigned long           li_state;
1798         struct list_head        li_request_list;
1799         struct mutex            li_list_mtx;
1800 };
1801
1802 struct ext4_li_request {
1803         struct super_block      *lr_super;
1804         struct ext4_sb_info     *lr_sbi;
1805         ext4_group_t            lr_next_group;
1806         struct list_head        lr_request;
1807         unsigned long           lr_next_sched;
1808         unsigned long           lr_timeout;
1809 };
1810
1811 struct ext4_features {
1812         struct kobject f_kobj;
1813         struct completion f_kobj_unregister;
1814 };
1815
1816 /*
1817  * This structure will be used for multiple mount protection. It will be
1818  * written into the block number saved in the s_mmp_block field in the
1819  * superblock. Programs that check MMP should assume that if
1820  * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
1821  * to use the filesystem, regardless of how old the timestamp is.
1822  */
1823 #define EXT4_MMP_MAGIC     0x004D4D50U /* ASCII for MMP */
1824 #define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
1825 #define EXT4_MMP_SEQ_FSCK  0xE24D4D50U /* mmp_seq value when being fscked */
1826 #define EXT4_MMP_SEQ_MAX   0xE24D4D4FU /* maximum valid mmp_seq value */
1827
1828 struct mmp_struct {
1829         __le32  mmp_magic;              /* Magic number for MMP */
1830         __le32  mmp_seq;                /* Sequence no. updated periodically */
1831
1832         /*
1833          * mmp_time, mmp_nodename & mmp_bdevname are only used for information
1834          * purposes and do not affect the correctness of the algorithm
1835          */
1836         __le64  mmp_time;               /* Time last updated */
1837         char    mmp_nodename[64];       /* Node which last updated MMP block */
1838         char    mmp_bdevname[32];       /* Bdev which last updated MMP block */
1839
1840         /*
1841          * mmp_check_interval is used to verify if the MMP block has been
1842          * updated on the block device. The value is updated based on the
1843          * maximum time to write the MMP block during an update cycle.
1844          */
1845         __le16  mmp_check_interval;
1846
1847         __le16  mmp_pad1;
1848         __le32  mmp_pad2[226];
1849         __le32  mmp_checksum;           /* crc32c(uuid+mmp_block) */
1850 };
1851
1852 /* arguments passed to the mmp thread */
1853 struct mmpd_data {
1854         struct buffer_head *bh; /* bh from initial read_mmp_block() */
1855         struct super_block *sb;  /* super block of the fs */
1856 };
1857
1858 /*
1859  * Check interval multiplier
1860  * The MMP block is written every update interval and initially checked every
1861  * update interval x the multiplier (the value is then adapted based on the
1862  * write latency). The reason is that writes can be delayed under load and we
1863  * don't want readers to incorrectly assume that the filesystem is no longer
1864  * in use.
1865  */
1866 #define EXT4_MMP_CHECK_MULT             2UL
1867
1868 /*
1869  * Minimum interval for MMP checking in seconds.
1870  */
1871 #define EXT4_MMP_MIN_CHECK_INTERVAL     5UL
1872
1873 /*
1874  * Maximum interval for MMP checking in seconds.
1875  */
1876 #define EXT4_MMP_MAX_CHECK_INTERVAL     300UL
1877
1878 /*
1879  * Function prototypes
1880  */
1881
1882 /*
1883  * Ok, these declarations are also in <linux/kernel.h> but none of the
1884  * ext4 source programs needs to include it so they are duplicated here.
1885  */
1886 # define NORET_TYPE     /**/
1887 # define ATTRIB_NORET   __attribute__((noreturn))
1888 # define NORET_AND      noreturn,
1889
1890 /* bitmap.c */
1891 extern unsigned int ext4_count_free(char *bitmap, unsigned numchars);
1892 void ext4_inode_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
1893                                 struct ext4_group_desc *gdp,
1894                                 struct buffer_head *bh, int sz);
1895 int ext4_inode_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
1896                                   struct ext4_group_desc *gdp,
1897                                   struct buffer_head *bh, int sz);
1898 void ext4_block_bitmap_csum_set(struct super_block *sb, ext4_group_t group,
1899                                 struct ext4_group_desc *gdp,
1900                                 struct buffer_head *bh);
1901 int ext4_block_bitmap_csum_verify(struct super_block *sb, ext4_group_t group,
1902                                   struct ext4_group_desc *gdp,
1903                                   struct buffer_head *bh);
1904
1905 /* balloc.c */
1906 extern void ext4_get_group_no_and_offset(struct super_block *sb,
1907                                          ext4_fsblk_t blocknr,
1908                                          ext4_group_t *blockgrpp,
1909                                          ext4_grpblk_t *offsetp);
1910 extern ext4_group_t ext4_get_group_number(struct super_block *sb,
1911                                           ext4_fsblk_t block);
1912
1913 extern void ext4_validate_block_bitmap(struct super_block *sb,
1914                                        struct ext4_group_desc *desc,
1915                                        unsigned int block_group,
1916                                        struct buffer_head *bh);
1917 extern unsigned int ext4_block_group(struct super_block *sb,
1918                         ext4_fsblk_t blocknr);
1919 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
1920                         ext4_fsblk_t blocknr);
1921 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
1922 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
1923                         ext4_group_t group);
1924 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
1925                                          ext4_fsblk_t goal,
1926                                          unsigned int flags,
1927                                          unsigned long *count,
1928                                          int *errp);
1929 extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
1930                                     s64 nclusters, unsigned int flags);
1931 extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *);
1932 extern void ext4_check_blocks_bitmap(struct super_block *);
1933 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
1934                                                     ext4_group_t block_group,
1935                                                     struct buffer_head ** bh);
1936 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
1937
1938 extern struct buffer_head *ext4_read_block_bitmap_nowait(struct super_block *sb,
1939                                                 ext4_group_t block_group);
1940 extern int ext4_wait_block_bitmap(struct super_block *sb,
1941                                   ext4_group_t block_group,
1942                                   struct buffer_head *bh);
1943 extern struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
1944                                                   ext4_group_t block_group);
1945 extern void ext4_init_block_bitmap(struct super_block *sb,
1946                                    struct buffer_head *bh,
1947                                    ext4_group_t group,
1948                                    struct ext4_group_desc *desc);
1949 extern unsigned ext4_free_clusters_after_init(struct super_block *sb,
1950                                               ext4_group_t block_group,
1951                                               struct ext4_group_desc *gdp);
1952 extern unsigned ext4_num_overhead_clusters(struct super_block *sb,
1953                                            ext4_group_t block_group,
1954                                            struct ext4_group_desc *gdp);
1955 ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
1956
1957 /* dir.c */
1958 extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
1959                                   struct file *,
1960                                   struct ext4_dir_entry_2 *,
1961                                   struct buffer_head *, char *, int,
1962                                   unsigned int);
1963 #define ext4_check_dir_entry(dir, filp, de, bh, buf, size, offset)      \
1964         unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
1965                                         (de), (bh), (buf), (size), (offset)))
1966 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
1967                                     __u32 minor_hash,
1968                                     struct ext4_dir_entry_2 *dirent);
1969 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
1970 extern int ext4_find_dest_de(struct inode *dir, struct inode *inode,
1971                              struct buffer_head *bh,
1972                              void *buf, int buf_size,
1973                              const char *name, int namelen,
1974                              struct ext4_dir_entry_2 **dest_de);
1975 void ext4_insert_dentry(struct inode *inode,
1976                         struct ext4_dir_entry_2 *de,
1977                         int buf_size,
1978                         const char *name, int namelen);
1979 static inline void ext4_update_dx_flag(struct inode *inode)
1980 {
1981         if (!EXT4_HAS_COMPAT_FEATURE(inode->i_sb,
1982                                      EXT4_FEATURE_COMPAT_DIR_INDEX))
1983                 ext4_clear_inode_flag(inode, EXT4_INODE_INDEX);
1984 }
1985 static unsigned char ext4_filetype_table[] = {
1986         DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
1987 };
1988
1989 static inline  unsigned char get_dtype(struct super_block *sb, int filetype)
1990 {
1991         if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FILETYPE) ||
1992             (filetype >= EXT4_FT_MAX))
1993                 return DT_UNKNOWN;
1994
1995         return ext4_filetype_table[filetype];
1996 }
1997
1998 /* fsync.c */
1999 extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
2000 extern int ext4_flush_unwritten_io(struct inode *);
2001
2002 /* hash.c */
2003 extern int ext4fs_dirhash(const char *name, int len, struct
2004                           dx_hash_info *hinfo);
2005
2006 /* ialloc.c */
2007 extern struct inode *__ext4_new_inode(handle_t *, struct inode *, umode_t,
2008                                       const struct qstr *qstr, __u32 goal,
2009                                       uid_t *owner, int handle_type,
2010                                       unsigned int line_no, int nblocks);
2011
2012 #define ext4_new_inode(handle, dir, mode, qstr, goal, owner) \
2013         __ext4_new_inode((handle), (dir), (mode), (qstr), (goal), (owner), \
2014                          0, 0, 0)
2015 #define ext4_new_inode_start_handle(dir, mode, qstr, goal, owner, \
2016                                     type, nblocks)                  \
2017         __ext4_new_inode(NULL, (dir), (mode), (qstr), (goal), (owner), \
2018                          (type), __LINE__, (nblocks))
2019
2020
2021 extern void ext4_free_inode(handle_t *, struct inode *);
2022 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
2023 extern unsigned long ext4_count_free_inodes(struct super_block *);
2024 extern unsigned long ext4_count_dirs(struct super_block *);
2025 extern void ext4_check_inodes_bitmap(struct super_block *);
2026 extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
2027 extern int ext4_init_inode_table(struct super_block *sb,
2028                                  ext4_group_t group, int barrier);
2029 extern void ext4_end_bitmap_read(struct buffer_head *bh, int uptodate);
2030
2031 /* mballoc.c */
2032 extern long ext4_mb_stats;
2033 extern long ext4_mb_max_to_scan;
2034 extern int ext4_mb_init(struct super_block *);
2035 extern int ext4_mb_release(struct super_block *);
2036 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
2037                                 struct ext4_allocation_request *, int *);
2038 extern int ext4_mb_reserve_blocks(struct super_block *, int);
2039 extern void ext4_discard_preallocations(struct inode *);
2040 extern int __init ext4_init_mballoc(void);
2041 extern void ext4_exit_mballoc(void);
2042 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
2043                              struct buffer_head *bh, ext4_fsblk_t block,
2044                              unsigned long count, int flags);
2045 extern int ext4_mb_alloc_groupinfo(struct super_block *sb,
2046                                    ext4_group_t ngroups);
2047 extern int ext4_mb_add_groupinfo(struct super_block *sb,
2048                 ext4_group_t i, struct ext4_group_desc *desc);
2049 extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
2050                                 ext4_fsblk_t block, unsigned long count);
2051 extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
2052
2053 /* inode.c */
2054 struct buffer_head *ext4_getblk(handle_t *, struct inode *,
2055                                                 ext4_lblk_t, int, int *);
2056 struct buffer_head *ext4_bread(handle_t *, struct inode *,
2057                                                 ext4_lblk_t, int, int *);
2058 int ext4_get_block_write(struct inode *inode, sector_t iblock,
2059                          struct buffer_head *bh_result, int create);
2060 int ext4_get_block(struct inode *inode, sector_t iblock,
2061                                 struct buffer_head *bh_result, int create);
2062 int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
2063                            struct buffer_head *bh, int create);
2064 int ext4_walk_page_buffers(handle_t *handle,
2065                            struct buffer_head *head,
2066                            unsigned from,
2067                            unsigned to,
2068                            int *partial,
2069                            int (*fn)(handle_t *handle,
2070                                      struct buffer_head *bh));
2071 int do_journal_get_write_access(handle_t *handle,
2072                                 struct buffer_head *bh);
2073 #define FALL_BACK_TO_NONDELALLOC 1
2074 #define CONVERT_INLINE_DATA      2
2075
2076 extern struct inode *ext4_iget(struct super_block *, unsigned long);
2077 extern int  ext4_write_inode(struct inode *, struct writeback_control *);
2078 extern int  ext4_setattr(struct dentry *, struct iattr *);
2079 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
2080                                 struct kstat *stat);
2081 extern void ext4_evict_inode(struct inode *);
2082 extern void ext4_clear_inode(struct inode *);
2083 extern int  ext4_sync_inode(handle_t *, struct inode *);
2084 extern void ext4_dirty_inode(struct inode *, int);
2085 extern int ext4_change_inode_journal_flag(struct inode *, int);
2086 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
2087 extern int ext4_can_truncate(struct inode *inode);
2088 extern void ext4_truncate(struct inode *);
2089 extern int ext4_punch_hole(struct file *file, loff_t offset, loff_t length);
2090 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
2091 extern void ext4_set_inode_flags(struct inode *);
2092 extern void ext4_get_inode_flags(struct ext4_inode_info *);
2093 extern int ext4_alloc_da_blocks(struct inode *inode);
2094 extern void ext4_set_aops(struct inode *inode);
2095 extern int ext4_writepage_trans_blocks(struct inode *);
2096 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
2097 extern int ext4_block_truncate_page(handle_t *handle,
2098                 struct address_space *mapping, loff_t from);
2099 extern int ext4_block_zero_page_range(handle_t *handle,
2100                 struct address_space *mapping, loff_t from, loff_t length);
2101 extern int ext4_zero_partial_blocks(handle_t *handle, struct inode *inode,
2102                              loff_t lstart, loff_t lend);
2103 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2104 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
2105 extern void ext4_da_update_reserve_space(struct inode *inode,
2106                                         int used, int quota_claim);
2107
2108 /* indirect.c */
2109 extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
2110                                 struct ext4_map_blocks *map, int flags);
2111 extern ssize_t ext4_ind_direct_IO(int rw, struct kiocb *iocb,
2112                                 const struct iovec *iov, loff_t offset,
2113                                 unsigned long nr_segs);
2114 extern int ext4_ind_calc_metadata_amount(struct inode *inode, sector_t lblock);
2115 extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks);
2116 extern void ext4_ind_truncate(handle_t *, struct inode *inode);
2117 extern int ext4_free_hole_blocks(handle_t *handle, struct inode *inode,
2118                                  ext4_lblk_t first, ext4_lblk_t stop);
2119
2120 /* ioctl.c */
2121 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
2122 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
2123
2124 /* migrate.c */
2125 extern int ext4_ext_migrate(struct inode *);
2126 extern int ext4_ind_migrate(struct inode *inode);
2127
2128 /* namei.c */
2129 extern int ext4_dirent_csum_verify(struct inode *inode,
2130                                    struct ext4_dir_entry *dirent);
2131 extern int ext4_orphan_add(handle_t *, struct inode *);
2132 extern int ext4_orphan_del(handle_t *, struct inode *);
2133 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
2134                                 __u32 start_minor_hash, __u32 *next_hash);
2135 extern int search_dir(struct buffer_head *bh,
2136                       char *search_buf,
2137                       int buf_size,
2138                       struct inode *dir,
2139                       const struct qstr *d_name,
2140                       unsigned int offset,
2141                       struct ext4_dir_entry_2 **res_dir);
2142 extern int ext4_generic_delete_entry(handle_t *handle,
2143                                      struct inode *dir,
2144                                      struct ext4_dir_entry_2 *de_del,
2145                                      struct buffer_head *bh,
2146                                      void *entry_buf,
2147                                      int buf_size,
2148                                      int csum_size);
2149
2150 /* resize.c */
2151 extern int ext4_group_add(struct super_block *sb,
2152                                 struct ext4_new_group_data *input);
2153 extern int ext4_group_extend(struct super_block *sb,
2154                                 struct ext4_super_block *es,
2155                                 ext4_fsblk_t n_blocks_count);
2156 extern int ext4_resize_fs(struct super_block *sb, ext4_fsblk_t n_blocks_count);
2157
2158 /* super.c */
2159 extern int ext4_calculate_overhead(struct super_block *sb);
2160 extern int ext4_superblock_csum_verify(struct super_block *sb,
2161                                        struct ext4_super_block *es);
2162 extern void ext4_superblock_csum_set(struct super_block *sb);
2163 extern void *ext4_kvmalloc(size_t size, gfp_t flags);
2164 extern void *ext4_kvzalloc(size_t size, gfp_t flags);
2165 extern void ext4_kvfree(void *ptr);
2166 extern int ext4_alloc_flex_bg_array(struct super_block *sb,
2167                                     ext4_group_t ngroup);
2168 extern const char *ext4_decode_error(struct super_block *sb, int errno,
2169                                      char nbuf[16]);
2170 extern __printf(4, 5)
2171 void __ext4_error(struct super_block *, const char *, unsigned int,
2172                   const char *, ...);
2173 #define ext4_error(sb, message...)      __ext4_error(sb, __func__,      \
2174                                                      __LINE__, ## message)
2175 extern __printf(5, 6)
2176 void ext4_error_inode(struct inode *, const char *, unsigned int, ext4_fsblk_t,
2177                       const char *, ...);
2178 extern __printf(5, 6)
2179 void ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t,
2180                      const char *, ...);
2181 extern void __ext4_std_error(struct super_block *, const char *,
2182                              unsigned int, int);
2183 extern __printf(4, 5)
2184 void __ext4_abort(struct super_block *, const char *, unsigned int,
2185                   const char *, ...);
2186 #define ext4_abort(sb, message...)      __ext4_abort(sb, __func__, \
2187                                                        __LINE__, ## message)
2188 extern __printf(4, 5)
2189 void __ext4_warning(struct super_block *, const char *, unsigned int,
2190                     const char *, ...);
2191 #define ext4_warning(sb, message...)    __ext4_warning(sb, __func__, \
2192                                                        __LINE__, ## message)
2193 extern __printf(3, 4)
2194 void ext4_msg(struct super_block *, const char *, const char *, ...);
2195 extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
2196                            const char *, unsigned int, const char *);
2197 #define dump_mmp_msg(sb, mmp, msg)      __dump_mmp_msg(sb, mmp, __func__, \
2198                                                        __LINE__, msg)
2199 extern __printf(7, 8)
2200 void __ext4_grp_locked_error(const char *, unsigned int,
2201                              struct super_block *, ext4_group_t,
2202                              unsigned long, ext4_fsblk_t,
2203                              const char *, ...);
2204 #define ext4_grp_locked_error(sb, grp, message...) \
2205         __ext4_grp_locked_error(__func__, __LINE__, (sb), (grp), ## message)
2206 extern void ext4_update_dynamic_rev(struct super_block *sb);
2207 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
2208                                         __u32 compat);
2209 extern int ext4_update_rocompat_feature(handle_t *handle,
2210                                         struct super_block *sb, __u32 rocompat);
2211 extern int ext4_update_incompat_feature(handle_t *handle,
2212                                         struct super_block *sb, __u32 incompat);
2213 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
2214                                       struct ext4_group_desc *bg);
2215 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
2216                                       struct ext4_group_desc *bg);
2217 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
2218                                      struct ext4_group_desc *bg);
2219 extern __u32 ext4_free_group_clusters(struct super_block *sb,
2220                                       struct ext4_group_desc *bg);
2221 extern __u32 ext4_free_inodes_count(struct super_block *sb,
2222                                  struct ext4_group_desc *bg);
2223 extern __u32 ext4_used_dirs_count(struct super_block *sb,
2224                                 struct ext4_group_desc *bg);
2225 extern __u32 ext4_itable_unused_count(struct super_block *sb,
2226                                    struct ext4_group_desc *bg);
2227 extern void ext4_block_bitmap_set(struct super_block *sb,
2228                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2229 extern void ext4_inode_bitmap_set(struct super_block *sb,
2230                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2231 extern void ext4_inode_table_set(struct super_block *sb,
2232                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
2233 extern void ext4_free_group_clusters_set(struct super_block *sb,
2234                                          struct ext4_group_desc *bg,
2235                                          __u32 count);
2236 extern void ext4_free_inodes_set(struct super_block *sb,
2237                                 struct ext4_group_desc *bg, __u32 count);
2238 extern void ext4_used_dirs_set(struct super_block *sb,
2239                                 struct ext4_group_desc *bg, __u32 count);
2240 extern void ext4_itable_unused_set(struct super_block *sb,
2241                                    struct ext4_group_desc *bg, __u32 count);
2242 extern int ext4_group_desc_csum_verify(struct super_block *sb, __u32 group,
2243                                        struct ext4_group_desc *gdp);
2244 extern void ext4_group_desc_csum_set(struct super_block *sb, __u32 group,
2245                                      struct ext4_group_desc *gdp);
2246 extern int ext4_register_li_request(struct super_block *sb,
2247                                     ext4_group_t first_not_zeroed);
2248
2249 static inline int ext4_has_group_desc_csum(struct super_block *sb)
2250 {
2251         return EXT4_HAS_RO_COMPAT_FEATURE(sb,
2252                                           EXT4_FEATURE_RO_COMPAT_GDT_CSUM |
2253                                           EXT4_FEATURE_RO_COMPAT_METADATA_CSUM);
2254 }
2255
2256 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
2257 {
2258         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
2259                 le32_to_cpu(es->s_blocks_count_lo);
2260 }
2261
2262 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
2263 {
2264         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
2265                 le32_to_cpu(es->s_r_blocks_count_lo);
2266 }
2267
2268 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
2269 {
2270         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
2271                 le32_to_cpu(es->s_free_blocks_count_lo);
2272 }
2273
2274 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
2275                                          ext4_fsblk_t blk)
2276 {
2277         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
2278         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
2279 }
2280
2281 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
2282                                               ext4_fsblk_t blk)
2283 {
2284         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
2285         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
2286 }
2287
2288 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
2289                                            ext4_fsblk_t blk)
2290 {
2291         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
2292         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
2293 }
2294
2295 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
2296 {
2297         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
2298                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
2299                         le32_to_cpu(raw_inode->i_size_lo);
2300         else
2301                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
2302 }
2303
2304 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
2305 {
2306         raw_inode->i_size_lo = cpu_to_le32(i_size);
2307         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
2308 }
2309
2310 static inline
2311 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
2312                                             ext4_group_t group)
2313 {
2314          struct ext4_group_info ***grp_info;
2315          long indexv, indexh;
2316          grp_info = EXT4_SB(sb)->s_group_info;
2317          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
2318          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
2319          return grp_info[indexv][indexh];
2320 }
2321
2322 /*
2323  * Reading s_groups_count requires using smp_rmb() afterwards.  See
2324  * the locking protocol documented in the comments of ext4_group_add()
2325  * in resize.c
2326  */
2327 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
2328 {
2329         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
2330
2331         smp_rmb();
2332         return ngroups;
2333 }
2334
2335 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
2336                                              ext4_group_t block_group)
2337 {
2338         return block_group >> sbi->s_log_groups_per_flex;
2339 }
2340
2341 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
2342 {
2343         return 1 << sbi->s_log_groups_per_flex;
2344 }
2345
2346 #define ext4_std_error(sb, errno)                               \
2347 do {                                                            \
2348         if ((errno))                                            \
2349                 __ext4_std_error((sb), __func__, __LINE__, (errno));    \
2350 } while (0)
2351
2352 #ifdef CONFIG_SMP
2353 /* Each CPU can accumulate percpu_counter_batch clusters in their local
2354  * counters. So we need to make sure we have free clusters more
2355  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
2356  */
2357 #define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
2358 #else
2359 #define EXT4_FREECLUSTERS_WATERMARK 0
2360 #endif
2361
2362 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
2363 {
2364         /*
2365          * XXX: replace with spinlock if seen contended -bzzz
2366          */
2367         down_write(&EXT4_I(inode)->i_data_sem);
2368         if (newsize > EXT4_I(inode)->i_disksize)
2369                 EXT4_I(inode)->i_disksize = newsize;
2370         up_write(&EXT4_I(inode)->i_data_sem);
2371         return ;
2372 }
2373
2374 struct ext4_group_info {
2375         unsigned long   bb_state;
2376         struct rb_root  bb_free_root;
2377         ext4_grpblk_t   bb_first_free;  /* first free block */
2378         ext4_grpblk_t   bb_free;        /* total free blocks */
2379         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
2380         ext4_grpblk_t   bb_largest_free_order;/* order of largest frag in BG */
2381         struct          list_head bb_prealloc_list;
2382 #ifdef DOUBLE_CHECK
2383         void            *bb_bitmap;
2384 #endif
2385         struct rw_semaphore alloc_sem;
2386         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
2387                                          * regions, index is order.
2388                                          * bb_counters[3] = 5 means
2389                                          * 5 free 8-block regions. */
2390 };
2391
2392 #define EXT4_GROUP_INFO_NEED_INIT_BIT           0
2393 #define EXT4_GROUP_INFO_WAS_TRIMMED_BIT         1
2394
2395 #define EXT4_MB_GRP_NEED_INIT(grp)      \
2396         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
2397
2398 #define EXT4_MB_GRP_WAS_TRIMMED(grp)    \
2399         (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2400 #define EXT4_MB_GRP_SET_TRIMMED(grp)    \
2401         (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2402 #define EXT4_MB_GRP_CLEAR_TRIMMED(grp)  \
2403         (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2404
2405 #define EXT4_MAX_CONTENTION             8
2406 #define EXT4_CONTENTION_THRESHOLD       2
2407
2408 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
2409                                               ext4_group_t group)
2410 {
2411         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
2412 }
2413
2414 /*
2415  * Returns true if the filesystem is busy enough that attempts to
2416  * access the block group locks has run into contention.
2417  */
2418 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
2419 {
2420         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
2421 }
2422
2423 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
2424 {
2425         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
2426         if (spin_trylock(lock))
2427                 /*
2428                  * We're able to grab the lock right away, so drop the
2429                  * lock contention counter.
2430                  */
2431                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
2432         else {
2433                 /*
2434                  * The lock is busy, so bump the contention counter,
2435                  * and then wait on the spin lock.
2436                  */
2437                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
2438                                   EXT4_MAX_CONTENTION);
2439                 spin_lock(lock);
2440         }
2441 }
2442
2443 static inline void ext4_unlock_group(struct super_block *sb,
2444                                         ext4_group_t group)
2445 {
2446         spin_unlock(ext4_group_lock_ptr(sb, group));
2447 }
2448
2449 /*
2450  * Block validity checking
2451  */
2452 #define ext4_check_indirect_blockref(inode, bh)                         \
2453         ext4_check_blockref(__func__, __LINE__, inode,                  \
2454                             (__le32 *)(bh)->b_data,                     \
2455                             EXT4_ADDR_PER_BLOCK((inode)->i_sb))
2456
2457 #define ext4_ind_check_inode(inode)                                     \
2458         ext4_check_blockref(__func__, __LINE__, inode,                  \
2459                             EXT4_I(inode)->i_data,                      \
2460                             EXT4_NDIR_BLOCKS)
2461
2462 /*
2463  * Inodes and files operations
2464  */
2465
2466 /* dir.c */
2467 extern const struct file_operations ext4_dir_operations;
2468
2469 /* file.c */
2470 extern const struct inode_operations ext4_file_inode_operations;
2471 extern const struct file_operations ext4_file_operations;
2472 extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
2473 extern void ext4_unwritten_wait(struct inode *inode);
2474
2475 /* inline.c */
2476 extern int ext4_has_inline_data(struct inode *inode);
2477 extern int ext4_get_inline_size(struct inode *inode);
2478 extern int ext4_get_max_inline_size(struct inode *inode);
2479 extern int ext4_find_inline_data_nolock(struct inode *inode);
2480 extern void ext4_write_inline_data(struct inode *inode,
2481                                    struct ext4_iloc *iloc,
2482                                    void *buffer, loff_t pos,
2483                                    unsigned int len);
2484 extern int ext4_prepare_inline_data(handle_t *handle, struct inode *inode,
2485                                     unsigned int len);
2486 extern int ext4_init_inline_data(handle_t *handle, struct inode *inode,
2487                                  unsigned int len);
2488 extern int ext4_destroy_inline_data(handle_t *handle, struct inode *inode);
2489
2490 extern int ext4_readpage_inline(struct inode *inode, struct page *page);
2491 extern int ext4_try_to_write_inline_data(struct address_space *mapping,
2492                                          struct inode *inode,
2493                                          loff_t pos, unsigned len,
2494                                          unsigned flags,
2495                                          struct page **pagep);
2496 extern int ext4_write_inline_data_end(struct inode *inode,
2497                                       loff_t pos, unsigned len,
2498                                       unsigned copied,
2499                                       struct page *page);
2500 extern struct buffer_head *
2501 ext4_journalled_write_inline_data(struct inode *inode,
2502                                   unsigned len,
2503                                   struct page *page);
2504 extern int ext4_da_write_inline_data_begin(struct address_space *mapping,
2505                                            struct inode *inode,
2506                                            loff_t pos, unsigned len,
2507                                            unsigned flags,
2508                                            struct page **pagep,
2509                                            void **fsdata);
2510 extern int ext4_da_write_inline_data_end(struct inode *inode, loff_t pos,
2511                                          unsigned len, unsigned copied,
2512                                          struct page *page);
2513 extern int ext4_try_add_inline_entry(handle_t *handle, struct dentry *dentry,
2514                                      struct inode *inode);
2515 extern int ext4_try_create_inline_dir(handle_t *handle,
2516                                       struct inode *parent,
2517                                       struct inode *inode);
2518 extern int ext4_read_inline_dir(struct file *filp,
2519                                 void *dirent, filldir_t filldir,
2520                                 int *has_inline_data);
2521 extern int htree_inlinedir_to_tree(struct file *dir_file,
2522                                    struct inode *dir, ext4_lblk_t block,
2523                                    struct dx_hash_info *hinfo,
2524                                    __u32 start_hash, __u32 start_minor_hash,
2525                                    int *has_inline_data);
2526 extern struct buffer_head *ext4_find_inline_entry(struct inode *dir,
2527                                         const struct qstr *d_name,
2528                                         struct ext4_dir_entry_2 **res_dir,
2529                                         int *has_inline_data);
2530 extern int ext4_delete_inline_entry(handle_t *handle,
2531                                     struct inode *dir,
2532                                     struct ext4_dir_entry_2 *de_del,
2533                                     struct buffer_head *bh,
2534                                     int *has_inline_data);
2535 extern int empty_inline_dir(struct inode *dir, int *has_inline_data);
2536 extern struct buffer_head *ext4_get_first_inline_block(struct inode *inode,
2537                                         struct ext4_dir_entry_2 **parent_de,
2538                                         int *retval);
2539 extern int ext4_inline_data_fiemap(struct inode *inode,
2540                                    struct fiemap_extent_info *fieinfo,
2541                                    int *has_inline);
2542 extern int ext4_try_to_evict_inline_data(handle_t *handle,
2543                                          struct inode *inode,
2544                                          int needed);
2545 extern void ext4_inline_data_truncate(struct inode *inode, int *has_inline);
2546
2547 extern int ext4_convert_inline_data(struct inode *inode);
2548
2549 /* namei.c */
2550 extern const struct inode_operations ext4_dir_inode_operations;
2551 extern const struct inode_operations ext4_special_inode_operations;
2552 extern struct dentry *ext4_get_parent(struct dentry *child);
2553 extern struct ext4_dir_entry_2 *ext4_init_dot_dotdot(struct inode *inode,
2554                                  struct ext4_dir_entry_2 *de,
2555                                  int blocksize, int csum_size,
2556                                  unsigned int parent_ino, int dotdot_real_len);
2557 extern void initialize_dirent_tail(struct ext4_dir_entry_tail *t,
2558                                    unsigned int blocksize);
2559 extern int ext4_handle_dirty_dirent_node(handle_t *handle,
2560                                          struct inode *inode,
2561                                          struct buffer_head *bh);
2562 #define S_SHIFT 12
2563 static unsigned char ext4_type_by_mode[S_IFMT >> S_SHIFT] = {
2564         [S_IFREG >> S_SHIFT]    = EXT4_FT_REG_FILE,
2565         [S_IFDIR >> S_SHIFT]    = EXT4_FT_DIR,
2566         [S_IFCHR >> S_SHIFT]    = EXT4_FT_CHRDEV,
2567         [S_IFBLK >> S_SHIFT]    = EXT4_FT_BLKDEV,
2568         [S_IFIFO >> S_SHIFT]    = EXT4_FT_FIFO,
2569         [S_IFSOCK >> S_SHIFT]   = EXT4_FT_SOCK,
2570         [S_IFLNK >> S_SHIFT]    = EXT4_FT_SYMLINK,
2571 };
2572
2573 static inline void ext4_set_de_type(struct super_block *sb,
2574                                 struct ext4_dir_entry_2 *de,
2575                                 umode_t mode) {
2576         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FILETYPE))
2577                 de->file_type = ext4_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
2578 }
2579
2580
2581 /* symlink.c */
2582 extern const struct inode_operations ext4_symlink_inode_operations;
2583 extern const struct inode_operations ext4_fast_symlink_inode_operations;
2584
2585 /* block_validity */
2586 extern void ext4_release_system_zone(struct super_block *sb);
2587 extern int ext4_setup_system_zone(struct super_block *sb);
2588 extern int __init ext4_init_system_zone(void);
2589 extern void ext4_exit_system_zone(void);
2590 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
2591                                  ext4_fsblk_t start_blk,
2592                                  unsigned int count);
2593 extern int ext4_check_blockref(const char *, unsigned int,
2594                                struct inode *, __le32 *, unsigned int);
2595
2596 /* extents.c */
2597 struct ext4_ext_path;
2598 struct ext4_extent;
2599
2600 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
2601 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
2602 extern int ext4_ext_index_trans_blocks(struct inode *inode, int extents);
2603 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
2604                                struct ext4_map_blocks *map, int flags);
2605 extern void ext4_ext_truncate(handle_t *, struct inode *);
2606 extern int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start,
2607                                  ext4_lblk_t end);
2608 extern void ext4_ext_init(struct super_block *);
2609 extern void ext4_ext_release(struct super_block *);
2610 extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
2611                           loff_t len);
2612 extern int ext4_convert_unwritten_extents(handle_t *handle, struct inode *inode,
2613                                           loff_t offset, ssize_t len);
2614 extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
2615                            struct ext4_map_blocks *map, int flags);
2616 extern int ext4_ext_calc_metadata_amount(struct inode *inode,
2617                                          ext4_lblk_t lblocks);
2618 extern int ext4_extent_tree_init(handle_t *, struct inode *);
2619 extern int ext4_ext_calc_credits_for_single_extent(struct inode *inode,
2620                                                    int num,
2621                                                    struct ext4_ext_path *path);
2622 extern int ext4_can_extents_be_merged(struct inode *inode,
2623                                       struct ext4_extent *ex1,
2624                                       struct ext4_extent *ex2);
2625 extern int ext4_ext_insert_extent(handle_t *, struct inode *,
2626                                   struct ext4_ext_path *,
2627                                   struct ext4_extent *, int);
2628 extern struct ext4_ext_path *ext4_ext_find_extent(struct inode *, ext4_lblk_t,
2629                                                   struct ext4_ext_path *);
2630 extern void ext4_ext_drop_refs(struct ext4_ext_path *);
2631 extern int ext4_ext_check_inode(struct inode *inode);
2632 extern int ext4_find_delalloc_range(struct inode *inode,
2633                                     ext4_lblk_t lblk_start,
2634                                     ext4_lblk_t lblk_end);
2635 extern int ext4_find_delalloc_cluster(struct inode *inode, ext4_lblk_t lblk);
2636 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
2637                         __u64 start, __u64 len);
2638
2639
2640 /* move_extent.c */
2641 extern void ext4_double_down_write_data_sem(struct inode *first,
2642                                             struct inode *second);
2643 extern void ext4_double_up_write_data_sem(struct inode *orig_inode,
2644                                           struct inode *donor_inode);
2645 void ext4_inode_double_lock(struct inode *inode1, struct inode *inode2);
2646 void ext4_inode_double_unlock(struct inode *inode1, struct inode *inode2);
2647 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
2648                              __u64 start_orig, __u64 start_donor,
2649                              __u64 len, __u64 *moved_len);
2650
2651 /* page-io.c */
2652 extern int __init ext4_init_pageio(void);
2653 extern void ext4_exit_pageio(void);
2654 extern void ext4_ioend_shutdown(struct inode *);
2655 extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
2656 extern ext4_io_end_t *ext4_get_io_end(ext4_io_end_t *io_end);
2657 extern int ext4_put_io_end(ext4_io_end_t *io_end);
2658 extern void ext4_put_io_end_defer(ext4_io_end_t *io_end);
2659 extern void ext4_io_submit_init(struct ext4_io_submit *io,
2660                                 struct writeback_control *wbc);
2661 extern void ext4_end_io_rsv_work(struct work_struct *work);
2662 extern void ext4_end_io_unrsv_work(struct work_struct *work);
2663 extern void ext4_io_submit(struct ext4_io_submit *io);
2664 extern int ext4_bio_write_page(struct ext4_io_submit *io,
2665                                struct page *page,
2666                                int len,
2667                                struct writeback_control *wbc);
2668
2669 /* mmp.c */
2670 extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
2671 extern void ext4_mmp_csum_set(struct super_block *sb, struct mmp_struct *mmp);
2672 extern int ext4_mmp_csum_verify(struct super_block *sb,
2673                                 struct mmp_struct *mmp);
2674
2675 /*
2676  * Note that these flags will never ever appear in a buffer_head's state flag.
2677  * See EXT4_MAP_... to see where this is used.
2678  */
2679 enum ext4_state_bits {
2680         BH_Uninit       /* blocks are allocated but uninitialized on disk */
2681          = BH_JBDPrivateStart,
2682         BH_AllocFromCluster,    /* allocated blocks were part of already
2683                                  * allocated cluster. */
2684 };
2685
2686 /*
2687  * Add new method to test whether block and inode bitmaps are properly
2688  * initialized. With uninit_bg reading the block from disk is not enough
2689  * to mark the bitmap uptodate. We need to also zero-out the bitmap
2690  */
2691 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
2692
2693 static inline int bitmap_uptodate(struct buffer_head *bh)
2694 {
2695         return (buffer_uptodate(bh) &&
2696                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
2697 }
2698 static inline void set_bitmap_uptodate(struct buffer_head *bh)
2699 {
2700         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
2701 }
2702
2703 /*
2704  * Disable DIO read nolock optimization, so new dioreaders will be forced
2705  * to grab i_mutex
2706  */
2707 static inline void ext4_inode_block_unlocked_dio(struct inode *inode)
2708 {
2709         ext4_set_inode_state(inode, EXT4_STATE_DIOREAD_LOCK);
2710         smp_mb();
2711 }
2712 static inline void ext4_inode_resume_unlocked_dio(struct inode *inode)
2713 {
2714         smp_mb();
2715         ext4_clear_inode_state(inode, EXT4_STATE_DIOREAD_LOCK);
2716 }
2717
2718 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
2719
2720 /* For ioend & aio unwritten conversion wait queues */
2721 #define EXT4_WQ_HASH_SZ         37
2722 #define ext4_ioend_wq(v)   (&ext4__ioend_wq[((unsigned long)(v)) %\
2723                                             EXT4_WQ_HASH_SZ])
2724 #define ext4_aio_mutex(v)  (&ext4__aio_mutex[((unsigned long)(v)) %\
2725                                              EXT4_WQ_HASH_SZ])
2726 extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
2727 extern struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
2728
2729 #define EXT4_RESIZING   0
2730 extern int ext4_resize_begin(struct super_block *sb);
2731 extern void ext4_resize_end(struct super_block *sb);
2732
2733 #endif  /* __KERNEL__ */
2734
2735 #endif  /* _EXT4_H */